Color image processor

BACKGROUND OF THE INVENTION

1. Field of the Invention

The present invention relates to a digital color image processor.

2. Description of Prior Art

In a digital color image processor, black characters in a document are usually discriminated to improve reproducibility of black characters. Characters (edges) are discriminated from achromatic image portions in a document by detecting an edge portion and an achromatic portion in the document. Various techniques are proposed for discriminating an achromatic image, and they discriminate an achromatic portion according to color data, that is, red (R), green (G) and blue (B) data. For example, in a technique, a difference of the maximum from the minimum of R, G and B color data, is used as chroma data. A portion having a larger difference is decided to be a color image, while a portion having a lower difference is decided to be an achromatic image.

It is also known to discriminate a dot area in a document and prevents Moire phenomenon by using the discrimination. In the discrimination, isolated dots isolated are detected from the gradation distribution of color data, and a dot area is detected according to the number of the dots in a unit pixel matrix.

On the other hand, in such an image processor, image forming mode such as the magnification or the density of an output image is changed automatically or manually by a user. The magnification can be changed by controlling resolution of color data. The density can be changed by controlling gradation characteristic such as background level.

When the gradation characteristic is changed for density control, the chroma data and the edge level are affected. For example, when the image reading conditions are set so as to remove the background of a document, it becomes difficult to decide an edge of a character having a low density. Further, when an image is reduced by changing magnification, a character having many lines therein is erroneously decided as dots. Thus, the quality of image reproduction is deteriorated.

SUMMARY OF THE INVENTION

An object of the present invention is to provide an image processor which improves the quality of a reproduced image surely by controlling image forming conditions when an image forming mode is set.

In one aspect of the invention, an image processor comprises an image scanner, and color data are obtained by reading a color image with a color image sensor. A specification device is provided to specify an image forming mode of the color image, and the image forming mode is changed according to the specified image forming mode when the color data is processed. The color data is changed according the image forming mode, and an area in the color data is discriminated for image processing of the color data with a reference value changed according to the image forming mode. Then, the color data is corrected according to the discrimination. For example, a gradation characteristic such as background level is specified as the image forming mode related to density, and a black portion in the color data is discriminated with the reference value changed according to the background level. In another example, magnification is specified as the image forming mode, and a dot area in the color data is discriminated with the reference value changed according to the magnification.

An advantage of the present invention is that a black portion in a color data can be discriminated surely when a gradation characteristic is changed.

Another advantage of the present invention is that a dot area in a color data can be discriminated surely even when magnification of the image is changed.

BRIEF DESCRIPTION OF THE DRAWINGS

These and other objects and features of the present invention will become clear from the following description taken in conjunction with the preferred embodiments thereof with reference to the accompanying drawings, and in which:

FIG. 1

is a sectional view of a digital color copying machine;

FIGS. 2A and 2B

are block diagrams of a signal processor;

FIG. 3

is a diagram of a basic picture in an operational panel;

FIG. 4

is a diagram for illustrating sampling in generating a histogram;

FIGS. 5 and 5A

illustrate various quantities obtained from the histogram;

FIG. 6

is a flowchart of automatic color selection;

FIG. 7

is a diagram for illustrating various quantities obtained from the histogram;

FIG. 8

is a flowchart of decision of document type;

FIG. 9

is a diagram for illustrating value signal and various signals G

25

-G

35

;

FIG. 10

is a block diagram of an HVC controller, an automatic exposure processor and a reverse HVC converter;

FIGS. 11 and 11A

are graphs of a distribution of value before and after automatic exposure on a monochromatic standard document;

FIGS. 12 and 12A

are graphs of a distribution of value before and after automatic exposure on a color standard document having white background; and

FIGS. 13A

,

13

B and

13

C are block diagrams of a region discriminator;

FIG. 14

is a diagram for explaining correction of phase shift due to color aberration;

FIG. 15

is a flowchart on the adjustment of background level;

FIG. 16

is a diagram of an example of adjustment of edge reference level;

FIG. 17

is a diagram of an example of adjustment of chroma reference level;

FIG. 18

is a diagram of a primary differential filter along the main scan direction;

FIG. 19

is a diagram of a primary differential filter along the subscan direction;

FIG. 20

is a diagram of a secondary differential filter;

FIG. 21A

is a graph of value distribution of five lines with different size from each other,

FIG. 21B

is a graph of primary differentials for the five lines, and

FIG. 21C

is a graph of secondary differentials for the five lines;

FIGS. 22 and 22A

illustrate an increase in chroma data W due to phase differences among R, G and B data, and WS obtained by smoothing;

FIG. 23

is a diagram of a smoothing filter;

FIG. 24

is a graph of a WREF table;

FIG. 25A

is a diagram an image consisting of cyan and magenta,

FIG. 25B

is a graph of image data of red, green and blue of the image shown in

FIG. 25A

, and

FIG. 25C

is a graph of chroma and color difference data for explaining erroneous detection of black at a boundary between cyan and yellow;

FIG. 26

is a diagram for showing two adjacent pixels along eight directions with respect to a pixel under interest (X) in filters for detecting white and black dot;

FIG. 27

is a diagram of four steps of reference levels for detecting dots and signals {overscore (AMI)}

0

-{overscore (AMI)}

3

;

FIG. 28

is a graph of a VMTF table;

FIGS. 29A and 29B

are block diagrams of an MTF correction section;

FIG. 30

is a timing chart of pixel clock, image data, driving voltage for laser diode, limit pulse, and driving voltage with a duty ratio;

FIG. 31

is a diagram of a Laplacian filter;

FIG. 32

is a graph of DMTF table;

FIG. 33

is a diagram of a smoothing filter for smoothing input data of 400 dpi to 300 dpi;

FIG. 34

is a diagram of a smoothing filter for smoothing input data of 400 dpi to 200 dpi;

FIG. 35

is a diagram of a smoothing filter for smoothing input data of 400 dpi to 100 dpi;

FIGS. 36A and 36B

are diagrams for explaining a slight extension of chromatic data outside a character and deletion of such extension;

FIGS. 37A and 37B

are diagrams of examples of images in correspondence to

FIGS. 36A and 36B

;

FIG. 38A

is a diagram of addition of correction data (hatched area) to an edge of an image, and

FIG. 38B

is a diagram of an amount of toners before correction (solid line) and after correction (dashed line);

FIG. 39

is a block diagram of a printer edge correction section;

FIGS. 40A

,

40

B and

40

C are diagrams of addition of PD

17-10

at a leading edge, at an intermediate point and at a trailing edge in an image;

FIG. 41

is a block diagram of a gamma correction section;

FIG. 42

is a graph of gamma correction table in value control mode;

FIG. 43

is a graph of gamma correction table in contrast control mode;

FIG. 44

is a graph of a relation of VIDEO

77-70

to VIDEO

47-40

for values of 1-7 of CO

2-0

; and

FIG. 45

is a graph of a relation of VIDEO

57-50

to VIDE

47-40

subtracted by background clearance data UDC

7-0

and corrected on slope by GDC

7-0

.

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

Referring now to the drawings, wherein like reference characters designate like or corresponding parts throughout the drawings, embodiments of the invention is described.

A. Structure of Digital Full Color Copying Machine

FIG. 1

shows a digital full color copying machine comprising an image scanner

30

, a data processor unit

10

and a printer section

20

. The image scanner

30

reads a document image, and the data processor unit

10

processes the data read received from the image scanner

30

. The printer section

20

print a full color or black image on a paper according to the data received from the data processor unit

10

. An outline of the digital copying machine is explained below.

In the image scanner

30

, a document is put on a platen glass

31

and covered with a plate

39

, or it is fed onto a platen

31

by an automatic document feeder (not shown) if mounted. A white plate

38

for shading correction is provided at an edge of the platen glass

31

. The document is exposed with a lamp

32

, and a light reflected from the document is guided through mirrors

33

a,

33

b

and

33

c

and converged by a lens

34

onto a color sensor

36

to be converted to color data of red (R), green (G) and blue (B). Then, they are sent to the data processor

10

. When the document image is read, a first slider

35

and a second slider

40

move at a speed of V and at a speed of V/2 mechanically by a motor

37

along the longitudinal direction (subscan direction) perpendicular to an electrical scan direction (main scan direction) of the color sensor

36

so that the entire document is scanned. The data processor

10

processes the image data electrically to output components of magenta (M), cyan (C), yellow (Y) and black (Bk) to the printer section

20

.

In the printer section

20

, the image signals of C, M, Y and Bk received from the data processor

10

are used to drive a laser diode

214

, and a laser beam emitted by the laser diode

214

propagates through a polygon mirror

215

, an f-θ lens

216

, mirrors

217

a

and

217

b

to expose a rotating photoconductor drum

206

charged beforehand by a charger

207

so as to form an electrostatic latent image. One of four development units

208

a,

208

b,

208

c

and

208

d

of toners of cyan, magenta, yellow and black is selected to develop the latent image with toners. On the other hand, a sheet of paper supplied from a cassette

201

a,

201

b

or

201

c

is carried by timing rollers

203

to be wound on a transfer drum

202

with an adsorption charger

204

. It is carried further to a transfer portion, and the toner image on the photoconductor drum

206

is transferred by a transfer charger

205

onto the sheet of paper. The above-mentioned printing process are repeated for four colors of yellow, magenta, cyan and black. That is, toner images of the four colors are transferred successively onto the sheet of paper. Then, the paper is separated by separation chargers

209

a,

209

b

and a claw

70

from the transfer drum

202

, passes through fixing rollers

210

a,

210

b

for fixing the toner image and is discharged onto a tray

211

.

B. Image Signal Processing

Next, image signal processing in the signal processor

10

is described.

FIGS. 2A and 2B

show image processing in the signal processor

10

. As explained above, the signal processor

10

receives analog image signals of 400 dots per inch of red, green and blue from the linear CCD sensor

36

on which a light reflected from a document is focused. In the A/D conversion section

100

, the analog image signals are converted to 8-bit digital data (256 gradation levels) of red (R), green (G) and blue (B). In order to eliminate scattering of reading of a quantity of light among CCD elements in the sensor

36

along a main scan direction for each of red, green and blue, a shading correction section

102

has stored reference data read on the white plate

38

in a memory (not shown), and when a document image is read, the data in the memory is converted to an inverted value thereof, and it is multiplied with a data on the document for shading correction. Next, a line correction section

104

adjusts the output of the data after shading correction according to positions of chips of red, green, and blue provided in the color CCD sensor

36

. A timing controller

106

controls timings for the CCD sensor

36

, the A/D conversion section

100

, the shading correction section

102

and the line correction section

104

. Then, the line correction section

104

sends the R, G and B data to line buffer

112

and magnification charge and move section

108

and a histogram generator

110

.

The magnification change and move section

108

has two linear memories, and magnification change and movement of data along a main scan direction along the CCD sensor

36

are controlled by changing timings of write and read to and from the memories.

The histogram generator

110

(

FIG. 4

) generates value signals from the R, G and B data obtained in a prescan to generate histograms. By using the histograms of the value signals, automatic color selection, background level and document mode are set automatically. The histogram generator

110

will be explained in detail later.

An HVC converter

114

converts the R, G and B data to value signals (V) and color difference signals (Cr and Cb). An edition processor

118

performs edition such as color change on the data received from the HVC converter

114

according to an instruction from an editor provided as an option.

On the other hand, an image interface

120

receives V, Cr and Cb data through an image selector

122

and sends the image data to an external equipment, or it received image data from the external equipment. In order to deal with various types of image data, the image interface

120

has a function to convert the V, Cr and Cb data to R, G and B signals, X, Y and Z signals, L*, a* and b* signals or the like, and vice versa. Further, C, M, Y and Bk data to be printed to the printer section

20

may be sent to the external equipment, and vice versa.

An image synthesis section

124

selects the V, Cr and Cb data received from the editor

118

or from the image selector

126

through the image interface

120

, and performs image synthesis of the data with another data received from the HVC converter

114

.

An HVC corrector

128

corrects the V, Cr and Cb data received from the image synthesis section

124

according to an instruction given with an operational panel

154

, in order to adjust image quality by a user in correspondence to three human senses of value (V), hue (H) and chroma (C).

An automatic exposure processor

130

controls background level of a document on value signals according to information obtained by the histogram generator

110

, as will be explained in detail later.

A reverse HVC converter

132

converts the V, Cr and Cb data again to R, G and B data.

In a color corrector

134

, a LOG corrector

136

converts the R, G and B data received from the reverse HVC converter

132

to density data DR, DG and DB, while a monochromatic data generator

138

generates value data from the R, G and B data in a color copy mode and generates gradation data DV for a monochromatic copy in a black copy mode. An undercolor/black paint section

140

calculates a difference between a maximum and a minimum of the density data DR, DG and DB as color information and a minimum among DR, DG and DB as a black component. The DR, DG and DB data are subtracted by the minimum to generate cyan, magenta and yellow data Co, Mo and Yo, while black data Bk is generated based on the minimum to be sent to a color data selector

144

. A masking operation section

142

converts the data Co, Mo and Yo to cyan, magenta and yellow data C, M and Y for color reproduction in the printer section

20

, and sends them to the color data selector

144

.

On the other hand, a region discriminator

146

discriminates a black character image, a dot image and the like, and generates a result (JD signal) and a correction signal (USM signal) based on the minimum MIN(R, G, B) and a difference between the maximum and the minimum (MAX(R, G, B)−MIN(R, G, B)). Further, a LIMOS signal is send to the printer section

20

to define a duty ratio of an output period to a pixel period. The output period means a period when a signal is output. The LIMOS signal is set to improve compatibility of reproduction of black characters and granularity of toner image.

An MTF corrector/sharpness controller

148

performs a various processing such as edge emphasis or smoothing on the data according to results obtained by the image discrimination section

146

for correcting a copy image appropriately.

A gamma correction/color balance section

150

controls a gamma curve (gradation correction curve) and color balance of C, M, Y and Bk data automatically or according to instruction given by the operational panel

154

. Then, the C, M, Y and Bk data and the LIMOS signal are sent to the printer section

20

.

A CPU

152

controls the signal processor

10

, and the operational panel

154

is used to give data and to display data.

C. Copy Mode

Next, copy modes of the full color copying machine are explained.

FIG. 3

shows a basic picture in the operational panel

154

, and a user can set various copy modes and the like.

First, background processing (automatic or manual) is explained. Background processing can be set as automatic exposure (AE) or manual setting wherein one of eight levels is set. In the automatic exposure, five following types of documents can be decided according to a histogram obtained by a prescan: color standard (white background, color background), color photograph, black-and-white standard, black-and-white photograph. Then, if the document type is decided to be a black-and-white standard document or a color standard (white background) document, value gradation correction to be explained later (

FIGS. 11 and 12

) is performed, otherwise a central level of the manual setting is selected as a default automatically (refer to Table 1). When manual setting is performed, contents as shown in Table 2 are set.

Next, document mode is explained (refer also to Table 1). A user can select automatic color selection (ACS) mode or one of four document modes. When ACS mode is selected, one of the four document modes is selected automatically according to the decision of document type based on prescan data, as shown in Table 1. When the document is decided to be a black-and-white document in the automatic color selection, black-and-white standard or photograph mode is selected and a copy is produced with a black copy mode. When the document is decided to be a color document, color standard or photograph mode is selected, and a copy is produced with a full color copy mode of four colors of cyan, magenta, yellow and black. When black-and-white standard or photograph mode is selected automatically or manually, the display in the operational panel is changed to a display for black-and-white mode (not shown), and a user selects a mixing ratio of R, G and B data as a document parameter in order to determine gradation data. (As a default data, average sensitivity distribution of red, green and blue is set for the ACS mode, while luminous efficiency is set as a default for the manual mode.) Further, a reproduction color can be selected among sixteen colors including black.

Though detailed explanation is omitted, a color resolution mode for reproducing C, M, B or Bk data in color copy mode or for reproducing Bk data in black copy mode for each document can be set.

D. Prescan and Generation of Histograms

In this embodiment, prescan is performed for determining document type and for performing automatic exposure (AE) and automatic color selection (ACS) according to the result of document type. The scan unit

35

in the image scanner

30

is positioned near the white plate

38

for shading correction opposite to a document reference position for a normal scan, in order to shorten a first copy time. When the start button in the operational panel

154

is pressed, the light source

32

is turned on, and the scanner

30

scans the white plate

38

and scans a document to generate histogram data thereof. Then, it returns to the document reference position. As will be explained later, automatic exposure (AE) and automatic color selection (ACS) are determined according to the generated histogram data, and a normal scan is started.

Next, generation of histograms in a prescan is explained.

FIG. 4

shows the histogram generator

110

which generates histograms in a document area in a prescan. The histogram generator

110

has first and second histogram memories

202

and

204

, and before a prescan, the two histogram memories

202

and

204

are initialized by writing “0” thereto at addresses of gradation levels of 0-255. In the histogram generator

110

, a value generator

200

receives the 8-bit R, G and B data and converts them to a value signal VH according to a following equation to be sent as an address signal to the first and second histogram memories

202

and

204

:

VH=

0.31640625*

R+

0.65625*

G+

0.02734375*

B

(1)

The value signal VH obtained resembles human sensitivity for observing an object. As explained above, the value data VH is used instead of the R, G and B data because the value data are used in the automatic exposure processing, as will be explained later.

A sampling interval circuit

206

determines intervals (a thinning out ratio) for storing data in the histogram memories

202

and

204

. This sampling is performed to reduce a memory capacity for prescan. If a histogram of all dots in a maximum document size of A3 is generated, a memory capacity of 32 megabits is needed. Then, in order to reduce the memory capacity to 1 megabits, data are sampled in this example for every eight dots along the main scan direction and for every four dots along the subscan direction for a document

31

, as shown in FIG.

5

. In

FIG. 5A

, dots denoted with circles are sampled in an effective document area represented with hatching.

A document size has been detected before a prescan, and the sampling interval circuit

206

receives various signals for sampling from the timing controller

106

. Among the signals, signals {overscore (HD)} and {overscore (VD)} are generated in a document area along the main scan direction and along the subscan direction. Then, the sampling interval circuit

206

allows generation of a histogram only in the document area determined by the signals {overscore (HD)} and {overscore (VD)}. A signal {overscore (TG)} denotes a synchronization clock signal along the main scan direction, and it is generated for each line. A signal VCLK denotes a synchronization clock signal of image data.

As to the histogram memories

202

and

204

, a read modify write cycle is performed for a period of eights dots. An address ADR of the histogram memory

202

,

204

corresponds to a value data (value gradation level), while a data at the address represents a frequency at the gradation level. When an address ADR is sent to the histogram memories

202

,

204

, data (frequency) at the address are read, and they are added by adders

208

,

210

by one. The sums are written to the histogram memories

202

,

204

at the same address. After a prescan is completed, the CPU

152

reads gradation data from the histogram memories

202

and

204

for various processings such as automatic exposure and automatic color selection to be explained later.

Two histogram memories

202

and

204

are used for automatic color selection and for document type determination. It is noted that data on all the dots can be written to the memory

202

because the {overscore (WE)} input of the first histogram memory

202

is always kept at L level. Thus, the first histogram memory

202

is used to generate a value histogram for a document simply. On the other hand, the second one

204

generates a histogram of achromatic dots in the document. In order to detect an achromatic dot, a minimum circuit

212

and a maximum circuit

214

detect a minimum (MIN) and a maximum (MAX) of input R, G and B data, and a subtraction circuit

216

calculates a difference between them. Then, a comparator

218

compares the difference (MAX−MIN) with a reference level SREF, and if the difference is smaller than the reference level, a data is allowed to be written to the second histogram memory

204

.

Automatic color selection and document type determination are performed based on first and second histograms generated in the first and second histogram memories

202

and

204

. As explained above, the histograms are generated on the value signals sampled in the effective document area; h

1

(

n

) denotes a frequency data at a value level n of the first histogram generated by the first histogram memory

202

, while h

2

(

n

) denotes a frequency data at a value level n of achromatic dots in the second histogram generated by the second histogram memory

204

.

Many quantities can be derived from the two histograms (h

1

(

n

) and h

2

(

n

)). Further, the CPU

152

generates a third histogram h

3

(

n

)=h

1

(

n

)−h

2

(

n

) by subtracting a frequency h

2

(

n

) of the second histogram memory

204

from a frequency h

1

(

n

) of the first histogram memory

202

. The third histogram represents a histogram for chromatic dots in a document. As shown in

FIG. 6

, several quantities can be obtained from the histograms h

1

(

n

) and h

3

(

n

). A sum W is obtained for levels n between μ1 and 255 from h

1

(

n

), and it represents a number of white dots, where a “dot” denotes each area detected by a linear sensor

36

in a document. That is, W denotes a dot number of the white background in a document. A sum M is obtained for levels n between μ2 and μ1 from h

1

(

n

), and it represents a number of dots of half-tone (grey) regions. A sum B is obtained for levels n between 0 and μ2 from h

1

(

n

), and it represents a number of dots in black areas. A sum C is obtained for levels n between τ2 and τ1 from h

3

(

n

) because dots of chromatic colors are counted.

\begin{matrix} \begin{matrix} W = \sum_{n = μ 1}^{255} h1 (n), \\ M = \sum_{n = μ 2}^{μ 1} h1 (n), \\ B = \sum_{n = 0}^{μ 2} h1 (n), \\ S = \sum_{n = 0}^{255} h1 (n), and \\ C = \sum_{n = σ 2}^{σ 1} h3 (n) . \end{matrix} & (2) \end{matrix}

In the automatic color selection, the sum C represents a number of dots in a color area in the document obtained from the second histogram h

3

(

n

).

In the document type determination, as shown in

FIG. 9

, frequency sums G

25

−G

20

and G

35

−G

30

are obtained in six value ranges from the two histograms h

2

(

n

) and h

3

(

n

).

E. Automatic Color Selection (ACS)

In the automatic color selection mode, a document put on the platen

31

is discriminated to be a black-and-white document or a color document to determine a copy mode automatically. Then, a color document is subjected to an image forming process of four colors (color copy mode). On the other hand, a black-and-white document is subjected to an image forming process of only black toners (black copy mode), and a copy speed is improved. Especially, when an automatic document feeder is used, even if black-and-white documents and color documents are placed in a mixed way, appropriate copying conditions can be set without manual operation by a user.

In the automatic color selection, a document type (black-and-white document or color document) is determined according to a ratio of dots of achromatic color to dots of chromatic color in a document. In concrete, a ratio of a sum C of chromatic dots to a sum S of the total dots obtained from the histogram is used to determine color copy or black copy. As explained above, C denotes a dot number in the color area obtained from h

3

(

n

), and S denotes a total number of dots in a document size. Then, the ratio C/S represents a ratio of chromatic color to (chromatic color plus achromatic color). If C/S is smaller than a reference value SREF, black copy mode is selected because there is if any a small part of chromatic color, while if C/S is larger than the reference value SREF, color copy mode is selected because there is a large part of chromatic color. By using S as a denominator, an effect of document size can be neglected in the automatic color selection.

FIG. 7

shows a flowchart of color selection of the CPU

152

. First, the histogram generator

110

generates histograms of value signal in the first and second histogram memories

202

and

204

(step S

100

). Next, C and S are obtained from the first and second histograms in the memories

202

and

204

(step S

102

), and a ratio C/S is calculated (step S

104

). If the ratio C/S is larger than the reference value SREF (YES at step S

104

), color copy mode is set (step S

108

), otherwise black copy mode is set (step S

110

).

E. Determination of Document Type

At a set-up stage of automatic exposure (AE), the CPU

152

determines five document types listed below, according to the information in the histogram memories

202

and

204

and the result of the automatic color selection (ACS) (refer to Table 1).

(1) Black-and-white photograph document: A black-and-white photograph, a black-and-white very precise dot image print, and the like.

(2) Black-and-white standard document: Documents with black characters, black liner images and the like, having a relatively white background.

(3) Color photograph document: A color silver salt photograph, a color image print with very fine dots and the like.

(4) Color standard document (with white background): A document having a relatively white background, including color characters, color linear images.

(5) Color standard document (with colored background): A document having a colored background.

The document type determination is also based on the histograms (refer to FIG.

8

). Concept of document type determination is explained first. As explained above on the automatic color selection, a color document and a black-and-white document are decided according to the ratio of chromatic dots to achromatic dots. That is, if the ratio is larger than the reference value, the document is decided to be a color document, otherwise it is decided to be a black-and-white document. Further, it is decided from the histograms if the document is a photograph document or a standard one. The standard document denotes a document including mainly characters, and a histogram thereof has a bi-level type distribution, as shown in

FIGS. 11 and 12

, having frequency peaks near levels 0 and 255. A document which have a colored background is also dealt with in this determination. If the histogram has a bi-level type distribution, the document is decided to be a standard document, otherwise it is decided to be a photograph document. In concrete, a number of dots in a density range at the white side is compared with that except the white side, and if the former is larger than a threshold value, the histogram is decided to have a bi-level type distribution, and the document is decided to be a standard document. This decision is effective for a black-and-white document and a color document having a white background. As to a color document, a standard document having a colored background and a color photograph document have to be discriminated. Then, if the histogram has an average distribution over a wide range, the document is decided to be a color photograph document, otherwise it is decided to be a color standard having a colored background. In concrete, a value signal level (0-255) is divided into a plurality of blocks (for example, six blocks for achromatic dots and for chromatic dots in FIG.

9

), and frequency sums in the blocks except one or more blocks near the highest value signal are obtained. Then, a difference between the maximum and the minimum among the sums is calculated, and if the difference is smaller than a threshold value, the document is decided to be a photograph document. Otherwise the document is decided to be a document having a colored background.

FIG. 8

shows a flowchart of determination of document type by the CPU

152

. First, various sums G

25

, G

24

, G

23

, G

22

, G

21

, G

20

, G

35

, G

34

, G

33

, G

32

, G

31

and G

30

defined below are calculated from frequency data h

2

(

n

) and h

3

(

n

) for achromatic dots and for chromatic dots. Further, a background level “a” and a character level “b” are calculated (step S

200

). The background level “a” represents a gradation level at the maximum frequency below 0.4 of output data ID in the second histogram memory

204

, and the character level “b” represents a gradation level at the maximum frequency above 0.6 of output data ID in the second histogram memory

204

(refer to examples shown in FIGS.

11

and

12

).

\begin{matrix} \begin{matrix} G_{25} = \sum_{n = 200}^{255} h2 (n), \\ G_{24} = \sum_{n = 128}^{199} h2 (n), \\ G_{23} = \sum_{n = 80}^{127} h2 (n), \\ G_{22} = \sum_{n = 48}^{79} h2 (n), \\ G_{21} = \sum_{n = 24}^{47} h2 (n), \\ G_{20} = \sum_{n = 0}^{23} h2 (n), \\ G_{35} = \sum_{n = 200}^{255} h3 (n), \\ G_{34} = \sum_{n = 128}^{199} h3 (n), \\ G_{33} = \sum_{n = 80}^{127} h3 (n), \\ G_{32} = \sum_{n = 48}^{79} h3 (n), \\ G_{31} = \sum_{n = 24}^{47} h3 (n), and \\ G_{30} = \sum_{n = 0}^{23} h3 (n) . \end{matrix} & (3) \end{matrix}

As shows in the left side of

FIG. 9

, levels 0-255 of value VH corresponds to the output data ID. The sums G

25

, G

24

, G

23

, G

22

, G

21

, G

20

, G

35

, G

34

, G

33

, G

32

, G

31

and G

30

correspond to sums obtained in six ranges of ID of 0.2 or below, 0.2-0.4, 0.4-0.6, 0.6-0.8, 0.8-1.1, 1.1 or above for the histogram h

2

(

n

) and the histogram h

3

(

n

). The sums G

25

, G

24

, G

23

, G

22

, G

21

, G

20

are obtained when a value data of MAX−MIN is smaller than the reference value SREF, otherwise the sums G

35

, G

34

, G

33

, G

32

, G

31

and G

30

are obtained. It is to be noted in

FIG. 9

that ranges shown with legends C, M, Y, R, G and B denote regions wherein VH data exist for colors of cyan, magenta, yellow, red, green and blue.

Next, in order to discriminate the above-mentioned five document types, a photograph document and a document having a colored background are decided. First, it is decided according to the result of the automatic color selection described above if the document is a color document (document types (3)-(5)) or a black-and-white document (document types (1)-(2) (step S

202

). If the result of the automatic color selection is a color document (YES at step S

202

), the flow proceeds to step S

204

to discriminate color documents (3)-(5) with a reference number α2. Then, if a ratio of a sum of frequencies of achromatic dots of 0.4 or more of output data ID and those of chromatic dots of 0.2 or more thereof to the sum S of total dots is smaller than is α2 (YES at step S

204

), the document is decided to be a color standard document having a colored background (step S

206

). The values of 0.4 or 0.2 correspond to ranges of various colors shown in FIG.

9

. Alternately, the former sum or the numerator in the ratio may be replaced with dots in the white background or a sum of frequencies not summed in the former sum. Then, as to image processing, automatic exposure is set, the document mode is set as a color standard mode, black character discrimination is set, and a gradation correction curve for the mode is set (step S

208

). On the other hand, if the ratio is not smaller than α2 (NO at step S

204

), it is decided further if a ratio of frequency sums of chromatic colors in a particular frequency block is very large or not (step S

210

). In concrete, a difference between a maximum and a minimum of sums in frequency blocks G

30

-G

34

for chromatic dots to the sum S of total dots is calculated, and if the ratio is not smaller than a reference value α3 (NO at step S

210

), the document is decided to be a color standard document having a colored background (step S

216

) because the image data is not average over all value gradation levels. Then, as to image processing, standard manual exposure is adopted and the exposure level is set at the center as a default value for manual setting, the document mode is set as a color standard mode, black character discrimination is set and a gradation correction curve for the mode is set (step S

218

). On the other hand, if the ratio is smaller than a reference value α3 (YES at step S

210

), the document is decided to be a color photograph document (step S

212

) because the image data is average over all value gradation levels. Then, as to image processing, photograph manual exposure is adopted and exposure level is set at the center as a default, the document mode is set as a color photograph mode, black character discrimination is not set, and a gradation correction curve is not changed (step S

214

).

If the result of the automatic color selection is decided not a color document (NO at step S

202

), the flow proceeds to step S

220

to discriminate color documents (1)-(3). with a reference number α1. Then, if a ratio of a sum of frequencies of achromatic dots of 0.4 or more of output data ID to the sum S of total dots is smaller than α1 (YES at step S

220

), the document is decided to be a black-and-white photograph document (step S

222

). Then, as to image processing, photograph manual exposure is adopted and the level is set at the center as a default value, the document mode is set as a black photograph mode, black character discrimination is not set, and a gradation correction curve is not changed (step S

224

). On the other hand, if the ratio is not smaller than is α1 (NO at step S

220

), the document is decided to be a black-and-white standard document (step S

226

). Then, as to image processing, automatic exposure is set, the document mode is set as a black standard mode, black character discrimination is not set, and a gradation correction curve for the mode is changed (step S

228

).

Finally, the result of decision of document type is displayed in the basic operation picture (refer to

FIG. 3

) of the operational panel

154

(step S

230

). Because document type is displayed in the operational panel

154

, a user can recognize the result of decision readily. If this were not displayed, a user may become uneasy.

In the processing explained above, five document types (1)-(5) can be decided, and image processing is set according thereto. Table 1 shows contents of automatic color selection (ACS), image processing and document mode, and Table 2 shows background processing in the modes.

TABLE 1

Document type and image processing

Black

characters

Grada-

Document

discrimi-

tion

Document

type

ACS

Background

nation

change

mode

Color

Color

Manual

Yes

Yes

Color

standard

(Standard)

standard

(colored

Center

back-

ground)

Color

Color

AE

No

Yes

Color

standard

standard

(white

back-

ground)

Color

Color

Manual

Yes

No

Color

photo-

(Photo-

photo-

graph

graph)

graph

Center

Black-

Black

AE

No

Yes

Black

and-

standard

white

standard

Black-

Black

Manual

No

No

Black

and-

(Photo-

Photo-

white

graph)

graph

photo-

Center

graph

TABLE 2

Background processing

Black: Vout = 256* (Vin-8-b)/{(a-8) -b}

AE

Color: Vout = 256* (Vin-8)/(a-8)

Manual

+2

Color standard:

Vout = 256* (Vin-8)/(256-8)

level

Black standard:

Vout = 256* (Vin-16)/(256-16)

Photograph:

Vout = 256* (Vin-8)/(256-8)

+1

Color standard:

Vout = 256* (Vin-8)/(240-8)

Black standard:

Vout = 256* (Vin-16)/(240-16)

Photograph:

Vout = 256* (Vin-8)/(244-8)

±0

Color standard:

Vout = 256* (Vin-8)/(224-8)

Black standard:

Vout = 256* (Vin-16)/(224-16)

Photograph:

Vout = 256* (Vin-8)/(232-8)

−1

Color standard:

Vout = 256* (Vin-8)/(208-8)

Black standard:

Vout = 256* (Vin-16)/(208-16)

Photograph:

Vout = 256* (Vin-8)/(220-8)

−2

Color standard:

Vout = 256* (Vin-8)/(192-8)

Black standard:

Vout = 256* (Vin-16)/(192-16)

Photograph:

Vout = 256* (Vin-8)/(208-8)

−3

Color standard:

Vout = 256* (Vin-8)/(176-8)

Black standard:

Vout = 256* (Vin-16)/(176-16)

Photograph:

Vout = 256* (Vin-8)/(196-8)

−4

Color standard:

Vout = 256* (Vin-8)/(160-8)

Black standard:

Vout = 256* (Vin-16)/(160-16)

Photograph:

Vout = 256* (Vin-8)/(184-8)

−5

Color standard:

Vout = 256* (Vin-8)/(144-8)

Black standard:

Vout = 256* (Vin-16)/(144-16)

Photograph:

Vout = 256* (Vin-8)/(176-8)

In an example of the determination of document type described above, automatic color selection is performed first to determine a color document and a black-and-white document. Then, a standard document (including a document with a white background and with a colored background) and a photograph document are decided next. However, the determination of document type may be performed in a different way. For example, if the existence of the background is determined according a frequency sum in the same value range is adopted in a step in correspondence to steps S

204

and

220

, it is possible to determine a standard document (having a white background) and a photograph document independently of the determination of a color document or a black-and-white document.

F. HVC Conversion and HVC Control

The copying machine of the embodiment processes image data by using conversion of data of red (R), green (G) and blue (B) to HVC data. The HVC converter

114

has a matrix operator for converting the R, G and B data to value signal (V) and two kinds of color difference signals (Cr and Cb).

\begin{matrix} \begin{matrix} (\begin{matrix} \begin{matrix} V \\ Cr \end{matrix} \\ Cb \end{matrix}) = (\begin{matrix} 0.31640625 & 0.65625 & 0.02734375 \\ 1 & - 0.9609375 & - 0.0390625 \\ - 0.32421875 & - 0.67578125 & 1 \end{matrix}) * \\ (\begin{matrix} \begin{matrix} V \\ Cr \end{matrix} \\ Cb \end{matrix}) . \end{matrix} & (4) \end{matrix}

Three attributes of value, chroma and hue of image are obtained as follows by using the signals V, Cr and Cb:

Value=V,

Chroma=(Cr

2

+Cb

2

)

½

,

and

Hue=arc tan (Cb/Cr). (5)

The conversion to signals V, Cr and Cb is used to improve image quality by adopting processing similar to human sense and to facilitate processing needed later such as image synthesis, automatic exposure and HVC adjustment.

The HVC converter

114

send signals to the edition processor

118

as well as to the image synthesis section

124

. The edition processor

118

performs edition of image such as color change. In the image synthesis section

124

, the signals received from the HVC converter

114

are stored once in a delay memory

116

so as to synchronize them with image signals received from the edition processor

118

. Then, the image synthesis section

124

synthesizes the output data V, Cr and Cb received from the delay memory

116

with the output data V, Cr and Cb of the edition processor

118

received through the image selector

126

. Image synthesis performed by the image synthesis section

124

includes for example synthesis by adding the two images and synthesis of characters.

The HVC correction unit

128

shown in

FIG. 10

is provided for controlling image quality. In order to control image for each of receive signals H, V and C, the HVC correction unit

128

comprises a matrix operator

128

a

for performing a matrix operation as shown below.

\begin{matrix} (\begin{matrix} \begin{matrix} V \\ Cr \end{matrix} \\ Cb \end{matrix}) = (\begin{matrix} 1 & 0 & 0 \\ 0 & q * \cos θ & - q * \cos θ \\ 0 & q * \sin θ & q * \cos θ \end{matrix}) * (\begin{matrix} \begin{matrix} V \\ Cr \end{matrix} \\ Cb \end{matrix}) . & (6) \end{matrix}

wherein “q” denotes a chroma control coefficient and θ denotes a hue control coefficient. These coefficients are output by the HVC correction unit

128

wherein they are selected among eight groups of coefficients according to 3-bit M data signal set by a user with the operational panel

154

. Thus, the image can be corrected by a user.

H. Automatic Exposure (AE)

Background processing is explained as an example of a processing using document type determined in a prescan. In this embodiment, value signal VH which resembles human luminous efficiency (brightness) is generated. Then, histograms of the value signal VH are generated, as explained before, and document type is decided by using the histograms. The background can be processed automatically according to the document type without changing color balance of a full color document. It can also be processed appropriately without special adjustment for a document including color and black images therein by discriminating the areas of the color and black images in a prescan. Image signals R, G and B are once converted to signals V, Cr and Cb, and automatic exposure is performed on the signals V, Cr and Cb. Then, the signals V, Cr and Cb are converted again to image signals R, G and B. Thus, the background level can be adjusted appropriately in the automatic exposure be determining it uniquely both on the full color mode and on the black mode. In the full color mode, because color component signals Cr and Cb are not subjected to any processing, color balance is not affected by the automatic exposure.

In concrete, background adjustment is performed by automatic exposure or by setting an exposure level manually. In the basic picture shown in

FIG. 3

, a user can select automatic exposure or a manual setting of an exposure level among eight levels. In the automatic exposure, five document types are discriminated according to document histogram information obtained in a prescan of a document, as explained above. As shown in

FIG. 8

, if the document type is color standard document (with white background) or black-and-white document, value gradation level is corrected as shown in

FIGS. 11 and 12

, while as to the other three document types, the center level of manual setting is set automatically as a default level (refer to Table 1).

In the automatic exposure section

130

, background is deleted according to document type determined with the histogram generator

110

. As to two document types of color standard document (white background) and black-and-white document, a look up table (AE table)

131

a

on value signal is used to obtain a value signal V

out

after automatic exposure bases on an input value signal V

in

before automatic exposure according to correction formulas explained below. That is, as to document type of black-and-white document,

V

out

=256*(V

in

−b−8)/{(a−8)−b}, (7)

and, as to document type of color standard document (with white background),

V

out

=256*(V

in

−8)/(a−8), (8)

wherein “a” denotes background level and “b” denotes character level. As to a black-and-white document, background is removed, and characters with light densities are made darker, as shown in

FIGS. 11 and 11A

. That is, value signal between a+8 to b is expanded in a wider range between 0 and 255 so as to delete input signals V

in

below a+8 and above b. On the other hand, as to a color document with white background, only the background is removed. That is, as shown in

FIGS. 12 and 12A

, value signal between 8 to b is expanded in a wider range between 0 and 255 so as to delete input signals V

in

above b. In this example, levels for deleting background are set at 0−8.

The background level “a” and the character level “b” are determined as described below. By using the value histogram h

1

(

n

) on the entire document obtained in the first histogram memory

202

, a gradation level “m” having the maximum frequency h

1

(

m

) is determined in a range of n=136-255 (or ID of 0.4 or less) . Then, “a” is set as m−8, and background value is set as 255. The histogram distribution around the level “m” will have a normal distribution with a scattering, and the scattering is set as ±8 in this example to delete gradations around the level “m” surely by setting “a” as m−8. Similarly, only for a black-and-white document, a gradation level “l” having the maximum frequency h

1

(

l

) is determined in a range of n=0-120 (or ID of 0.4 or more). Then, “b” is set as l+8, and value of a character is set as 0. Gradation around the level “l” can be made black surely by setting “b” as l+8. As to a color standard document with white background, “b” is not used for automatic exposure because characters may not necessarily be black.

Because color difference signals Cr and Cb or color information are not changed (or D

in

=D

out

in tables

131

b

and

131

c

) in the automatic exposure section

130

, only darkness (V) is controlled, and color balance is not affected. That is, background level is controlled automatically without changing color information of a color document.

In the manual setting mode using the operational panel

154

, value signal is corrected to change background level. As shown in

FIG. 3

, seven levels from +2 to +1 and from −1 to −5 can be set manually at dark and light sides with respect to the center level (0). The steps at positive (dark) side means enhancement of background, while those at negative (light) side means deletion of background. Table 2 shows examples of background processing at each level from +2 to −5 for mode of the three document types: color standard mode, black standard mode and photograph mode.

When a document includes a color area and a black area, it is desirable that gradation expression after background processing is continuous between the areas. Then, beside the histogram analysis, features of input image are extracted in a prescan to determine image areas in a document. In an image area of color standard document type, gradation correction is performed according to Eq. (9), and in an image area of black-and-white standard document type, gradation correction is performed according to Eq. (8). Because Eqs. (8) and (9) are similar to each other, gradation expression becomes smooth even if background is processed differently in the two types of areas.

The black character discrimination (color blur) processing by the MTF correction section

148

is another image processing performed in correspondence to document type. As shown in Table 1, this processing is performed on a color standard document in order to optimize image reproduction of black characters when a color image and a black-and-white image are included in a document. First, in an area on which the area discrimination section

146

decides to be an edge, the data of cyan, magenta and yellow are decreased, while the data of black is increased for edge emphasis to broaden the characters somewhat by adding a value data V to 100% of the black data Bk.

The reverse HVC converter

132

converts the signals V, Cr and Cb again to signals R, G and B by using a following an inverse matrix operation:

\begin{matrix} (\begin{matrix} \begin{matrix} R \\ G \end{matrix} \\ B \end{matrix}) = (\begin{matrix} 1 & 0.68359375 & 0 \\ 1 & - 0.328125 & - 0.0390625 \\ 1 & 0 & 0.97265625 \end{matrix}) * (\begin{matrix} \begin{matrix} V \\ Cr \end{matrix} \\ Cb \end{matrix}) . & (9) \end{matrix}

After this reverse conversion, data processing can be performed on data of three primary colors.

G. Region Discrimination

(G-1) Correction of Phase Shift

FIGS. 13A-13C

show the region discriminator

146

. In a part shown in

FIG. 13A

, pseudo-chroma data W

7-0

and pseudo-value data V

7-0

used for discrimination are generated from the R, G and B data received from the reverse HVC converter

132

. The pseudo-chroma data W

7-0

is generated as a difference of the maximum from the minimum, or, MAX(R, G, B)−MIN(R, G, B), of the R, G and B data. That is, a color image having high chroma has a large W, while an achromatic or monochromatic image having low chroma has a small W.

In the part shown in

FIG. 13A

, phase shift due to color aberration of the R, G and B data are corrected before determining the chroma data. Red light has longer wavelength components, and blue light has shorter wavelength components. As shown in

FIG. 14

, color aberration of an optical system arises at two ends along the main scan direction. When a vertical line existing at the center of the image is read, as shown at the center in

FIG. 14

, no color aberration occurs. However, when vertical lines existing near the two ends of the image are read, as shown at the right and left in

FIG. 14

, color aberration occurs at the ends of the lens

34

. The light R of longer wavelengths is converged at an inner side, while light B of shorter wavelengths is converged at an outer side, in contrast to the line image at the central portion of the lens. Thus, phases of R, G and B of an image are shifted on the color sensor

36

. Color aberration causes color shifts at edges of a character image. Especially, an edge of a black character is liable to be discriminated erroneously. Then, for example, colors may extend around a character, or a character may be cut into parts.

In order to correct the phase shift, phase correctors

1461

-

1464

are provided as shown in

FIG. 13A

to correct four kinds of color aberration states. For example, the phase corrector

1461

outputs ¼*R(n+1)+¾*R(n) for an n-th red data R(n) and ¾*B(n)+¼*B(n−1) for an n-th blue data B(n). The other phase correctors

1462

-

1464

have different predetermined correction coefficients for replacing with R(n) and B(n). The correction of the R, G and B color data for color aberration is shown below. (1) For a shift of ¼ dot at the reference side in the main scan direction,

R

(

n

)=0.25

*R

(

n+

1)+0.75

*R

(

n

),

G(n)=G(n),

and

B

(

n

)=0.75

*B

(

n

)+0.25

*B

(

n

−1). (10)

(2) For a shift of ⅛ dot at the reference side in the main scan direction,

R

(

n

)=0.125

*R

(

n

+1)+0.875

*R

(

n

),

G(n)=G(n),

and

B

(

n

)=0.875

*B

(

n

)+0.125

*B

(

n

−1). (11)

(3) At the center in the main scan direction,

R(n)=R(n),

G(n)=G(n),

and

B(n)=B(n). (12)

(4) For a shift of ⅛ dot at the opposite side to the reference side in the main scan direction,

R

(

n

)=0.875

*R

(

n

)+0.125

*R

(

n

−1),

G(n)=G(n),

and

B

(

n

)=0.125

*B

(

n

+1)+0.875

*B

(

n

). (13)

(5) For a shift of ¼ dot at the opposite side to the reference side in the main scan direction,

R

(

n

)=0.75

*R

(

n

)+0.25

*R

(

n

−1),

G(n)=G(n),

and

B

(

n

)=0.25

*B

(

n

+1)+0.75

*B

(

n

). (14)

In

FIG. 13A

, chroma detectors

1465

-

1469

calculate pseudo-chroma data which is equal to a difference of the maximum from the minimum of R, G and B data of the phase correctors

1461

-

1464

. Then, a data selector

1471

selects the minimum data among the differences to output it as the pseudo-chroma data corrected for color aberration.

This correction is based on that there are no phase shifts of R, G and B lights when color aberration is corrected and that {MAX(R, G, B)−MIN(R, G, B)} will become minimum when there are no phase shifts of R, G and B. According to the color aberration characteristics, the direction of the phase shift of the R sensor is opposite to that of the B sensor, as shown in

FIG. 14

, and the amount of the phase shift is about the same for the R and B sensors. Then, at the reference side in the main scan direction, R data is shifted by +1/n dot, and B data is shifted by −1/n dot where n=4 or 8. On the other hand, at the opposite side to the reference side in the main scan direction, R data is shifted by −1/n dot, and B data is shifted by +1/n dot. Five chroma data for different n's are calculated, and the data nearest to the achromatic color among them is selected as the chroma data W.

On the other hand, a pseudo-value data generator

1470

generates the minimum data MIN(R, G, B) as the pseudo-value data V

7-0

. Because the minimum data MIN(R, G, B) is used as the pseudo-value data V

7-0

, the dependence on the colors of a document can be vanished as to the determination of a black character, an edge in a dot image or an isolated dot. The color in R, G and B data having the minimum data corresponds to a color component having the highest density among them. Therefore, the color has gradation level characteristic similar to colors such as yellow having a high value and to black or colors such as blue having a low value. Therefore, an edge or an isolated point can be detected without affected by the chroma and hue in contrast to the processing using the original value data. By using the corrected pseudo-chroma data, scattering of read data due to the lens can be neglected.

(G-2) Adjustment of Reference Levels

FIG. 15

shows a flowchart for automatic adjustment of the reference level for area discrimination for black edge and dot area according to image forming conditions specified by a user. First, the flow branches according to the background level (step S

300

). According as the background level is +2, +1, 0, −1 or −2, the adjustment value AX for background level is set at

256

,

240

,

224

,

208

or

144

(step S

301

-S

308

).

Then, edge reference ER is set as ER*(AX/

256

) (step S

309

). Thus, the edge reference ER (such as EDGREF

17-10

, EDGREF

27-20

, EDGREF

37-30

or EDGREF

47-40

for the comparators

1521

-

1524

shown in

FIG. 13C

) for detecting an edge in an image is changed with the automatic exposure level or the background value set manually described above on the automatic exposure.

FIG. 16

shows an example of the control of the edge reference level ER. This as due to a fact that an edge of a character is determined by a line contrast relative to the white reference level (gradation level 255). Needless to say, the edge reference for the secondary differential filter can also be changed automatically.

Next, the WREF table (

1513

in

FIG. 13B

) for determining the reference level for black character discrimination is also controlled automatically. The reference BR(V) or WREF for a WREF table as a function of pseudo-value data V of MIN(R, G, B) is a reference level for chroma data when the background level is 255. The BR(V) is set as BR(V)*(V*

256

/AX) for the background level AX (step S

310

).

FIG. 17

shows an example of the reference level control of the WREF table.

Next, it is decided if RX*RY is equal to or less than 1 (step S

311

), where RX and RY represent magnifications in the main scan direction and in the subscan direction. In other words, it is decided whether the image is reduced or not. When the image is reduced (YES at step S

311

), the reference level Cntref

27-20

for a comparator

1554

(

FIG. 13C

) used for binarization for the number of isolated points (refer to

FIG. 13C

) is changed to copy magnification. That is, if the reference level Cntref

27-20

=CX, Cntref

27-20

is set as CX/(RX*RY) (step S

312

) to be set for the comparator

1554

for deciding the number of isolated points. If the image is not reduced (NO at step S

311

), the level CX is not changed (step S

312

). When an image is enlarged, it becomes difficult to detect dots. However, when an image is enlarged, the resolution for reading is increased, and a Moire pattern tends not to be liable to happen even if this correction is not operated. Therefore, it is not needed to change the level CX.

FIGS. 13B and 13C

are block diagrams of the region discriminator

146

which discriminates black character areas and dot image areas in a document image. The discrimination of black characters comprises four steps of (a) detection of a character (edge), (b) detection of black pixel, (c) detection of a region which is liable to be detected as black, and (d) generation of black edge reproduction signal which is performed by the MTF corrector

148

. The first to third steps are explained below in detail.

(G-3) Detection of Character (Edge)

First, detection of a character (edge) is explained in detail. A character has two elements of edge parts and uniform parts interposed by edge parts. If a character is thin, it has only edge portions. Then, the existence of a character is decided by detecting edges.

In the region discriminator

146

shown in

FIG. 13A

, the value signal V

7-0

generated by the HVC converter

114

is received through a negative/positive inverter

1501

to a line memory

1502

. If {overscore (NEGA)} signal set by an operator with the operational panel is L level, the inverter

1501

inverts the input data.

The data in the line memory is sent to primary differential filters

1503

and

1504

shown in

FIGS. 18 and 19

for the main scan direction and for the subscan direction each having a 5*5 matrix and to a secondary differential filter

1508

shown in FIG.

20

. In this embodiment, edges are detected with two kinds of differential filter because each has a feature.

FIG. 21A

shows value (lightness) distribution of five lines with different size from each other. Further,

FIG. 21B

shows primary differentials for the five lines, and

FIG. 21C

shows secondary differentials for the five lines. The primary differential filter outputs a higher detection value than the secondary one at an edge of a thick line (of a width of four pixels or larger). That is, the primary differential filter is suitable for detecting a thick edge of a width of four pixels or larger, while the secondary differential filter is suitable for detecting a thin edge of a width less than four pixels. In the region discriminator

146

, an edge of a character is detected if at least one of the primary and secondary filters outputs a value larger than a threshold value. Then, the detection precision of edge can be maintained irrespective of a width of a line.

The primary differential filters

1503

and

1504

along the main scan direction and along the subscan direction receive data read from the line memory

1502

. The obtained differentials are sent to absolute value circuits

1505

and

1506

to obtain absolute values thereof. The absolute values are needed because the primary differential filters

1403

and

1504

have negative coefficients. Then, an operator

1507

receives the absolute values and outputs an average FL

17-10

thereof. The average is used to take two differentials along the two directions into account. The average FL

17-10

of the first differentials is sent to comparators

1521

,

1523

,

1525

and

1527

for edge decision.

The secondary differential filter

1508

receives data from the line memory

1502

and an obtained second differential D

7-0

is output to an absolute value circuit

1509

to output an absolute value FL

27-20

thereof. The absolute value is needed because the secondary differential filter

1408

also have negative coefficients. The absolute value FL

27-20

of the secondary differential is sent to comparators

1522

,

1524

,

1526

and

1528

for edge decision. The secondary differential D

7-0

is also sent to a VMTF table

1512

shown in FIG.

28

. The VMTF table

1512

outputs value edge component VMTF

7-0

in correspondence to the secondary differential D

7-0

.

The comparator

1521

for edge decision shown in

FIG. 13C

compares the first differential FL

17-10

with a first edge reference level EDGREF

17-10

, and it outputs a signal of L level if the first differential FL

17-10

is larger than the first edge reference level EDGREF

17-10

. On the other hand, the comparator

1522

for edge decision compares the second differential FL

27-20

with a second edge reference level EDGREF

27-20

, and it outputs a signal of L level if the second differential FL

27-20

is larger than the second edge reference level EDGREF

27-20

. An AND gate

1533

receives the results of the comparison by the comparators

1521

,

1522

and it outputs an {overscore (EG)} signal if a signal of L level is received from at least one of the comparators

1521

and

1522

. The {overscore (EG)} signal means an edge.

(G-4) Decision of Black Pixel

Next, decision of black pixel is explained in detail. Black is detected based on chroma W

7-0

, or if the chroma W

7-0

is smaller than a reference value, the pixel is decided as black. However, the value of chroma W

7-0

may become high for a black pixel. For example, when the image sensor

14

vibrates when the image is read, the phases of data of red, green and blue may shift slightly relative to each other, as shown at a graph in FIG.

22

. In this case, the chroma W

7-0

becomes large as shown in another graph in FIG.

22

A. If the pixel is decided if the chroma W

7-0

is smaller than a reference value, the pixel is erroneously decided as a color pixel. Then, in this embodiment, erroneous decision can be prevented by smoothing the chroma data before the decision. That is, the chroma data W

7-0

is first received from the HVC converter

114

by another line memory

1514

, and it is smoothed by a filter

1515

of 3*3 matrix shown in FIG.

23

. Chroma data WS

7-0

after smoothing has a more gradual value, as shown in the lower part in FIG.

22

. Then, the above-mentioned type of erroneous decision can be prevented.

A comparator

1529

receives the chroma data WS

7-0

and compares it with a chroma reference data WREF

7-0

. If the chroma data WS

7-0

is smaller than the chroma reference data WREF

7-0

, the pixel is decided to be black, and the comparator

1529

sends {overscore (BK)} signal to an AND gate

1537

. As shown in

FIG. 24

, the chroma reference data WREF

7-0

is determined by the WREF table

1513

according to the value data V

7-0

. The WREF table

1513

has a feature that if the value data V

7-0

is larger than a predetermined value, WREF

7-0

is decreased linearly with the value V

7-0

. This takes into account that black pixels determined erroneously will become evident. (

FIG. 17

also shows the WREF table, but

FIG. 24

shows another example of the WREF table.) The AND gate

1537

outputs {overscore (BKEG)} which means an edge of a black pixel if the pixel is a pixel at an edge ({overscore (EG)}=L), it is a black pixel ({overscore (BK)}=L) and {overscore (BKEGEN)}=L.

(G-5) Decision of a Region Liable to be Detected as Black Character

Next, the detection of a region which is liable to be detected as black character is explained in detail. If only the detection of a character (edge) and the detection of black pixel mentioned above are performed, a character having a low value V

7-0

and a low chroma WS

7-0

such as dark blue and deep green is liable to be decided erroneously as an edge of a black character. Further, if a color and its complementary color, such as cyan and yellow, as shown in

FIG. 25A

, are adjacent to each other, and image data of red, green and blue are read as shown in

FIG. 25B

, the chroma WS

7-0

may become low at the boundary between them or change to black there, as shown in FIG.

25

C. Such a point is also liable to be decided erroneously as an edge of a black character. For example, such an erroneous decision may happen when a blue character is printed on a background of yellow.

In order to solve the problem, a uniform color part is detected in the embodiment. Then, even if the pixel is decided a black pixel, the decision is canceled if it is located in a region of uniform color part. Thus, a black character can be decided more precisely.

The uniform color part has features that it is not an edge, that it is a pixel in a color mode area and that a number of pixel having low value exceeds a certain number within a prescribed area. Then, the uniform color part is detected as follows: The comparators

1423

and

1524

decide that the outputs FL

17-10

and FL

27-20

of the primary and secondary differential filters are lower than third and fourth edge reference levels EDGREF

37-30

and EDREF

47-40

, an AND gate

1534

outputs signal {overscore (BETA

1

)} which means a pixel not existing at an edge. Further, if a comparator

1530

decides that the chroma data WS

7-0

is smaller than a reference value WREF

27-20

, it outputs a signal {overscore (COL)} which means a color data. Further, if a comparator

1531

decides that the value data V

17-10

is smaller than a reference value VREF

17-10

, it outputs a signal {overscore (VL

1

+L )}. Then, the AND gate

1538

receives the signals {overscore (BETA

1

)}, {overscore (COL)} and {overscore (VL

1

+L )} and outputs a signal {overscore (CAN)} which means that the pixel is not at an edge, that the pixel is in a color mode area and that the pixel has a low value. Then, the pixel is taken as a uniform part having a chromatic color not located in a background. A counter

1542

counts the number of the signals {overscore (CAN)} in the unit of 9*9 pixels. If the number Cntref

17-10

of the signals {overscore (CAN)} is smaller than a reference value Cntref

7-0

, a comparator

1542

outputs a signal {overscore (BKEGON)}.

An AND gate

1544

outputs the above-mentioned signal {overscore (BKEG)} delayed by a delay circuit

1541

and the above-mentioned signal {overscore (BKEGON)}. That is, even when the signal {overscore (BKEG)} on the decision of a black edge is received, if the signal {overscore (BKEGON)} is not received or if the pixel is located in a uniform color part, the decision of black edge is canceled, and the AND gate

1544

does not output a signal {overscore (PAPA)}. In other words, edge emphasis is performed only for a black character in a monochromatic background. On the other hand, the number of pixels of a uniform color part is less than the prescribed reference value, the decision of black edge is kept to be valid.

(G-6) Decision of Dot Area

Next, decision of dot area is explained in detail. Dot area means an area of an image composed of dots. As shown in

FIG. 13B

, the filters

1510

and

1511

for detection white dots and black dots receive data output from the line memory

1502

. Each filter decides if a pixel under interest is larger (white dots) or smaller (black dots) than a level AMIREF

7-0

along the all directions with respect to an average of two pixels surrounding the pixel under interest along eight directions, as shown in FIG.

26

. Further, if the pixel under interest is larger than the eight adjacent pixels, it is decided as a white dot ({overscore (WAMI)}=L), while if the pixel under interest is smaller than the eight adjacent pixels, it is decided as a black dot ({overscore (KAMI)}=L).

In concrete, the filter

1510

for detecting white dots shown in

FIG. 13B

outputs a signal {overscore (WAMI)} of L level when each condition of Eq. (16) is satisfied and each condition of Eq. (17) is satisfied. Further, the filter

1511

for detecting black dots shown in

FIG. 13B

also outputs a signal {overscore (KAMI)} of L level when each condition of Eq. (16) is satisfied and each condition of Eq. (17) is satisfied.

X−

(

a

11

+a

22

)/2>AMIREF

7-0

,

X−

(

a

31

+a

32

)/2>AMIREF

7-0

,

X−

(

a

51

+a

42

)/2>AMIREF

7-0

,

X−

(

a

53

+a

43

)/2>AMIREF

7-0

,

X−

(

a

55

+a

44

)/2>AMIREF

7-0

,

X−

(

a

35

+a

34

)/2>AMIREF

7-0

,

X−

(

a

15

+a

24

)/2>AMIREF

7-0

,

and

X−

(

a

13

+a

23

)/2>AMIREF

7-0

. (15)

X>a

22

,

X>a

32

,

X>a

42

,

X>a

43

,

X>a

44

,

X>a

34

,

X>a

24

,

and

X>a

23

. (16)

Further, the filter

1511

for detecting black dots shown in

FIG. 13B

also outputs a signal {overscore (KAMI)} of L level when each condition of Eq. (17) is satisfied and each condition of Eq. (18) is satisfied.

X−

(

a

11

+a

22

)/2<AMIREF

7-0

,

X−

(

a

31

+a

32

)/2<AMIREF

7-0

,

X−

(

a

51

+a

42

)/2<AMIREF

7-0

,

X−

(

a

53

+a

43

)/2<AMIREF

7-0

,

X−

(

a

55

+a

44

)/2<AMIREF

7-0

,

X−

(

a

35

+a

34

)/2<AMIREF

7-0

,

X−

(

a

15

+a

24

)/2<AMIREF

7-0

,

and

X−

(

a

13

+a

23

)/2<AMIREF

7-0

. (17)

X<a

22

,

X<a

32

,

X<a

42

,

X<a

43

,

X<a

44

,

X<a

34

,

X<a

24

,

and

X<a

23

. (18)

The counters

1550

and

1551

receive signals {overscore (WAMI)} and {overscore (KAMI)} output by the filters

1510

and

1511

, and they count a number of signals of L level in a 41*9 pixel matrix. The counts thereof are sent to a maximum detector

1552

which outputs a maximum thereof Amicnt

7-0

to four comparators

1553

-

1556

. The comparators

1553

-

1556

compare it with four steps of reference levels CNTREF

17-10

, CNTREF

27-20

, CNTREF

37-30

and CNTREF

47-40

to quantize it, and they output {overscore (AMIO)}, {overscore (AMI

1

)}, {overscore (AMI

2

)} and {overscore (AMI

3

)} if it is larger than the reference signals (refer to FIG.

27

).

(G-7) Other Types of Decision

The region discriminator

146

further decides some points explained below. A comparator

1532

is provided to decide a high light area. It compares the value data V

7-0

with a second reference level VREF

27-20

, and if the value data V

7-0

is larger than the second reference level VREF

27-20

, it outputs a signal {overscore (VH

1

)} which means that the pixel exists in a highlight area. The comparators

1527

and

1528

are provided to decide an area not located at an edge. They compare the first differential FL

17-10

and the second differential FL

27-20

with seventh and eighth reference levels EDGref

77-70

and EDGref

87-80

. If the first differential FL

17-10

and the second differential FL

27-20

are smaller than seventh and eighth reference levels EDGref

77-70

and EDGref

87-80

, a signal {overscore (BETA

2

)} which means a pixel not located at an edge is sent to an AND gate

1536

. The AND gate

1536

also receives the above-mentioned {overscore (VH

1

)} signal from the comparator

1531

, and it outputs a signal {overscore (HLIGHT)} which means a highlight area through a delay circuit

1546

.

The comparators

1525

and

1526

also receive the first differential FL

17-10

and the second differential FL

27-20

and compare them with fifth and sixth reference levels EDGref

57-50

and EDGref

67-60

. If the first differential FL

17-10

and the second differential FL

27-20

are larger than the reference levels EDGref

57-50

and EDGref

67-60

, signals of L level are sent to an NOR gate

1525

. If a signal is received from either of the comparators

1525

and

1526

, the NOR gate

1525

outputs a signal {overscore (EG

2

)} which means an edge highlight area through a delay circuit

1546

as a signal {overscore (MAMA)}.

H. MTF Corrector

FIGS. 29A and 29B

show block diagrams of the MTF corrector

148

which performs edge emphasis and smoothing most suitable for the image data VIDEO

7-0

and MVIDEO

7-0

received from the color corrector

134

according to the kind of pixels recognized by the signals ({overscore (AMIO)}-{overscore (AMI

3

)}, {overscore (MAMA)}, {overscore (PAPA)}, {overscore (EDG)} and {overscore (HLIGHT)}) and printing situation recognized by status signals ({overscore (MODE)}, {overscore (CMY)}/K, {overscore (BKER)}, {overscore (COLER)}). Further, a duty ratio of laser emission is changed according to the kind of image recognized by the region discriminator

146

. Still further, a prescribed value is added to pixel data at edges to correct amounts of excess or deficient toners.

The MTF corrector

148

recognizes the color of toners based on {overscore (CMY)}/K signal. If the signal is L level, toners of cyan, magenta or yellow is printed. It also recognizes one of following modes by using three signals {overscore (MODE)}, {overscore (BKER)} and {overscore (COLER)}: Full color standard mode ({overscore (BKER)}=H, {overscore (COLER)}=L and {overscore (MODE)}=H), full color photographic mode ({overscore (BKER)}=H, {overscore (COLER)}=H and {overscore (MODE)}=L), monochromatic color standard mode ({overscore (BKER)}=H, {overscore (COLER)}=L and {overscore (MODE)}=H), monochromatic color photograph mode ({overscore (BKER)}=H, {overscore (COLER)}=L and {overscore (MODE)}=L), monochromatic standard mode ({overscore (BKER)}=L, {overscore (COLER)}=L and {overscore (MODE)}=H), and monochromatic photographic mode ({overscore (BKER)}=L, {overscore (COLER)}=L and {overscore (MODE)}=L). Further, it recognizes the kind of a pixel to be printed by using the result of region discrimination as follows: A highlight region of uniform density ({overscore (HLIGHT)}=L), a non-edge region ({overscore (HLIGHT)}=H, {overscore (EDG)}=H, {overscore (PAPA)}=H), a color edge region ({overscore (HLIGHT)}=H, {overscore (EDG)}=L, {overscore (PAPA)}=H), and a black edge region ({overscore (HLIGHT)}=H, {overscore (EDG)}=L, {overscore (PAPA)}=L).

(H-1) Explanation of Various Modes

Before explaining the MTF corrector

148

, MTF correction in each mode mentioned above is explained. First, MTF correction in the full color standard mode ({overscore (MODE)}=H, {overscore (BKER)}=H and {overscore (COLER)}=L) is explained. Table 3 compiles signal levels of various signals received by a controller

1601

, printing situations represented by the levels and signals of DMPX

0

, DMPX

1

, DMPX

5

and DMPX

6

.

First, MTF correction of a pixel at a black edge ({overscore (HLIGHT)}=H, {overscore (EDG)}=L, {overscore (PAPA)}=L) is explained. When black toners are used for printing ({overscore (CMY)}/K=H), VIDEO

37-30

is obtained by adding edge component VMTF

7-0

of value to ordinary image data SD

7-0

for edge emphasis. The edge component VMTF

7-0

of value is used instead of an edge component DMTF

7-0

of density because the former is more sensitive than the latter on an edge due to background. If the pixel composes a dot image, the edge emphasis (or VMTF

7-0

) is limited according to the degree or density of dots. Thus, a Moire pattern is prevented to occur.

TABLE 3

Full color standard mode

{overscore (CMY)}/K

{overscore (HLIGHT)}

{overscore (EDG)}

{overscore (PAPA)}

DMPX1

DMPX0

USM

DMPX6

DMPX5

VIDEO

L

L

—

—

highlight

L

H

0

H

L

FSD

(CMY

H

H

H

non-edge

L

H

0

H

H

SD

mode)

H

L

H

color

H

H

DMTF

H

H

SD

edge

H

L

L

black

L

L

0

L

H

MIN

edge

H

L

—

—

highlight

L

H

0

H

L

FSD

(BK

H

H

H

non-edge

L

H

0

H

H

SD

mode)

H

L

H

color

L

H

0

H

H

SD

edge

H

L

L

black

H

L

VMTF

H

H

SD

edge

When cyan, magenta or yellow toners are used for printing ({overscore (CMY)}/K=L), edge emphasis is not performed, and a minimum data MIN

7-0

is obtained in a 5*5 or 3*3 matrix as VIDEO

37-30

. Then, a very narrow extended line at an edge as shown in

FIG. 36A

in an area represented with a dashed circle can be removed as shown in FIG.

36

B. By using the minimum data MIN

7-0

, image data can be decreased to zero only inside a black character. Then, the black character can be printed with edge emphases without white peripheral lines as shown in FIG.

37

A. If image data of cyan, magenta or yellow is subtracted by, for example, an edge detection quantity (such as FL

17-10

or FL

27-20

in this embodiment), white peripheral lines as shown in

FIG. 37A

are observed.

For a pixel in a color edge region ({overscore (HLIGHT)}=H, {overscore (EDG)}=L, {overscore (PAPA)}=H), edge emphases is not performed when black toners are used in printing, and ordinary pixel data SD

7-0

is used as VIDEO

37-30

. In other words, edge emphasis is not performed for an edge of a color character for black printing so that black fringe of a color character can be prevented. On the other hand, when cyan, magenta or yellow toners are used for printing, density edge component DTMF

7-0

is added to the ordinary pixel data SD

7-0

as VIDEO

37-30

.

For a pixel in a highlight region of uniform density ({overscore (HLIGHT)}=L), edge emphasis is not performed, and FSD

7-0

subjected to smoothing is used as image data VIDEO

37-30

. Then, noises in the highlight region becomes not noticeable.

For a pixel in a non-edge region ({overscore (HLIGHT)}=H, {overscore (EDG)}=H, {overscore (PAPA)}=H), edge emphasis is not performed and ordinary image data SD

7-0

is used as image data VIDEO

37-30

.

Next, MTF correction in the full color photographic mode ({overscore (BKER)}=H, {overscore (COLER)}=H and {overscore (MODE)}=L) is explained. Table 4 compiles signal levels of various signals received by the controller

1601

, printing situations represented by the levels and signals of DMPX

0

, DMPX

1

, DMPX

5

and DMPX

6

.

TABLE 4

Full color standard mode

{overscore (CMY)}/K

{overscore (HLIGHT)}

{overscore (EDG0)}

{overscore (PAPA)}

DMPX1

DMPX0

USM

DMPX6

DMPX5

VIDEO

L

L

—

—

highlight

L

H

0

H

L

FSD

(CMY

H

H

H

non-edge

L

H

0

H

L

FSD

mode)

H

L

H

color

H

H

DMTF

H

L

FSD

edge

H

L

L

black

H

H

DMTF

H

L

FSD

edge

H

L

—

—

highlight

L

H

0

H

L

FSD

(BK

H

H

H

non-edge

L

H

0

H

L

FSD

mode)

H

L

H

color

H

H

DMTF

H

L

FSD

edge

H

L

L

black

H

H

DMTF

H

L

FSD

edge

TABLE 5

Monochromatic color standard mode

{overscore (CMY)}/K

{overscore (HLIGHT)}

{overscore (EDG0)}

DMPX1

DMPX0

USM

DMPX6

DMPX5

VIDEO

—

L

—

highlight

L

H

0

H

L

FSD

H

H

non-edge

L

H

0

H

H

SD

L

H

L

CMY mode,

L

L

DMTF

H

H

SD

edge

H

H

L

BK mode,

L

H

0

H

H

SD

edge

For a pixel in a black edge region ({overscore (HLIGHT)}=H, {overscore (EDG)}=L, {overscore (PAPA)}=L) and in a color edge region ({overscore (HLIGHT)}=H, {overscore (EDG)}=L, {overscore (PAPA)}=H), edge emphasis is not performed when black toners are used in printing, and ordinary image data SD

7-0

is used as VIDEO

37-30

, while edge emphasis is performed when cyan, magenta or yellow toners are used in printing, by adding density edge component DMTF

7-0

to ordinary pixel data SD

7-0

to output the sum as VIDEO

37-30

. Thus, black fringe can be prevented.

For a pixel in a highlight region of uniform density ({overscore (HLIGHT)}=L), edge emphasis is not performed, and FSD

7-0

subjected to smoothing is used as image data VIDEO

37-30

. Then, noises in the highlight region becomes not noticeable.

For a pixel in a non-edge region ({overscore (HLIGHT)}=H, {overscore (EDG)}=H, {overscore (PAPA)}=H), edge emphasis is not performed and image data FSD

7-0

subjected to smoothing is used as image data VIDEO

37-30

.

Next, MTF correction in the monochromatic color photography mode ({overscore (BKER)}=H, {overscore (COLER)}=L and {overscore (MODE)}=L) is explained. Table 6 compiles signal levels of various signals received by the controller

1601

, printing situations represented by the levels and signals of DMPX

0

, DMPX

1

, MDMPX

5

and DMPX

6

.

TABLE 6

Monochromatic color photography mode

{overscore (CMY)}/K

{overscore (HLIGHT)}

{overscore (EDG0)}

DMPX1

DMPX0

USM

DMPX6

DMPX5

VIDEO

—

L

—

highlight

L

H

0

H

L

FSD

H

H

non-edge

L

H

0

H

L

FSD

L

H

L

CMY mode,

L

L

DMTF

H

L

FSD

edge

H

H

L

BK mode,

L

H

0

H

L

FSD

edge

For a pixel in a black edge region ({overscore (HLIGHT)}=H, {overscore (EDG)}=L, {overscore (PAPA)}=L) and in a color edge region ({overscore (HLIGHT)}=H, {overscore (EDG)}=L, {overscore (PAPA)}=H), edge emphases is performed only when cyan, magenta or yellow toners are used in printing, by adding density edge component DMTF

7-0

to FSD

7-0

subjected to smoothing to output the sum as VIDEO

37-30

so as not to deteriorate gradation characteristics of half-tone pixels. Thus, a black fringe of a color character can be prevented.

For a pixel in a highlight region of uniform density ({overscore (HLIGHT)}=L), edge emphasis is not performed, and FSD

7-0

subjected to smoothing is used as image data VIDEO

37-30

. Then, noises in the highlight region becomes not noticeable.

For a pixel in a non-edge region ({overscore (HLIGHT)}=H, {overscore (EDG)}=H, {overscore (PAPA)}=H), edge emphasis is not performed and image data FSD

7-0

subjected to smoothing is used as image data VIDEO

37-30

.

Next, MTF correction in the monochromatic standard mode ({overscore (BKER)}=L, {overscore (COLER)}=L and {overscore (MODE)}=H) is explained. Table 7 compiles signal levels of various signals received by the controller

1601

, printing situations represented by the levels and signals of DMPX

0

, DMPX

1

, MDMPX

5

and DMPX

6

.

TABLE 7

Monochromatic standard mode

{overscore (CMY)}/K

{overscore (HLIGHT)}

{overscore (EDG0)}

DMPX1

DMPX0

USM

DMPX6

DMPX5

VIDEO

—

L

—

highlight

L

H

0

H

L

FSD

H

H

non-edge

L

H

0

H

H

SD

L

H

L

CMY mode,

L

L

0

H

H

SD

edge

H

H

L

BK mode,

H

L

VMTF

H

H

SD

edge

For a pixel in a black edge region ({overscore (HLIGHT)}=H, {overscore (EDG)}=L, {overscore (PAPA)}=L) and in a color edge region ({overscore (HLIGHT)}=H, {overscore (EDG)}=L, L, {overscore (PAPA)}=H), edge emphasis is performed when black toners are used in printing, by adding value edge component VMTF

7-0

to ordinary pixel data SD

7-0

to output the sum as VIDEO

37-30

, while edge emphasis is not performed when cyan, magenta or yellow toners are used in printing, and ordinary image data SD

7-0

is used as VIDEO

37-30

.

For a pixel in a highlight region of uniform density ({overscore (HLIGHT)}=L), edge emphasis is not performed, and FSD

7-0

subjected to smoothing is used as image data VIDEO

37-30

. Then, noises in the highlight region becomes not noticeable.

For a pixel in a non-edge region ({overscore (HLIGHT)}=H, {overscore (EDG)}=H, {overscore (PAPA)}=H), edge emphasis is not performed and ordinary image data SD

7-0

is used as image data VIDEO

37-30

.

Finally, MTF correction in the monochromatic photography mode ({overscore (BKER)}=L, {overscore (COLER)}=L and {overscore (MODE)}=L) is explained. Table 8 compiles signal levels of various signals received by the controller

1601

, printing situations represented by the levels and signals of DMPX

0

, DMPX

1

, MDMPX

5

and DMPX

6

.

TABLE 8

Monochromatic photography mode

{overscore (CMY)}/K

{overscore (HLIGHT)}

{overscore (EDG0)}

DMPX1

DMPX0

USM

DMPX6

DMPX5

VIDEO

—

L

—

highlight

L

H

0

H

L

FSD

H

H

non-edge

L

H

0

H

L

FSD

L

H

L

CMY mode,

L

H

0

H

L

FSD

edge

H

H

L

BK mode,

H

H

DMTF

H

L

FSD

edge

For a pixel in a black edge region ({overscore (HLIGHT)}=H, {overscore (EDG)}=L, {overscore (PAPA)}=L) and in a color edge region ({overscore (HLIGHT)}=H, {overscore (EDG)}=L, {overscore (PAPA)}=H), edge emphases is performed by adding density edge component DMTF

7-0

to FSD

7-0

subjected to smoothing to output the sum as VIDEO

37-30

so as not to deteriorate gradation characteristics of half-tone pixels.

For a pixel in a highlight region of uniform density ({overscore (HLIGHT)}=L), and for a pixel in a non-edge region ({overscore (HLIGHT)}=H, {overscore (EDG)}=H, {overscore (PAPA)}=H), edge emphasis is not performed and image data FSD

7-0

subjected to smoothing is used as image data VIDEO

37-30

.

(H-2) MTF Correction

Next, MTF (mutual transfer) correction performed by the MTF corrector

148

shown in

FIGS. 29A and 29B

is explained. A controller

1601

for MTF correction parameters receives control signals {overscore (AMI

0

)}-{overscore (AMI

3

)}, {overscore (HLIGHT)}, {overscore (EDG)}, {overscore (PAPA)} and {overscore (MAMA)} from the region discriminator

146

. Further, the controller receives control signals {overscore (MODE)}, {overscore (CMY)}/K, {overscore (BKER)} and {overscore (COLER)}. The signal {overscore (MODE)} represents a kind of a document set by the key

78

in the operational panel, and it is set to be L level in the photography modes and H level in the standard modes. The signal {overscore (CMY)}/K is a status signal representing a printing situation, and it is set to be L level for printing with cyan, magenta or yellow toners and H level for printing with black toners. The signal {overscore (BKER)} requires signal processing in the monochromatic modes. The signal {overscore (COLER)} requires signal processing in the monochromatic color modes. The signals {overscore (BKER)} and {overscore (COLER)} are signals on a region. The controller

1601

supplies DMPX

0

-DMPX

6

shown in Tables 3-8 and a signal LIMOS shown in Table 9.

TABLE 9

Setting of duty ratio

MODE

{overscore (MAMA)}

{overscore (AMI0)}

LIMOS

H

L

—

L

—

L

L

H

H

H

L

—

—

H

The signal LIMOS changes a duty ratio of the laser diode emitting according to the image data. A period when the laser diode does not emit may be provided in one pixel clock cycle. In such a case, the duty ratio is defined as a ratio of the laser emission period in one pixel clock cycle.

FIG. 30

shows a timing chart on driving the laser diode wherein two types of a driving signal for the laser diode (LD) having duty ratios of 100% and 80% are shown. If the signal LIMOS=L, the duty ratio is set to be 100% in order to prevent a Moire pattern. If the signal LIMOS=H, the duty ratio is set to be 80% to reduce noises between lines along the main scan direction. If {overscore (MODE)}=H or the pixel is at an edge or in a dot in a cot image in the standard modes, the signal LIMOS is set to be L in order to improve the reproducibility at an edge and in a dot image. On the other hand, in the photography modes and at a non-edge region in the standard modes, the signal LIMOS=H to provide non-emitting periods in order to make noises between lines unnoticeable.

The signals {overscore (MODE)}, {overscore (CMY)}/K, {overscore (BKER)} and {overscore (COLER)} and an inverted signal of the signal {overscore (PAPA)} are also sent to a NAND gate

1602

. Then, the NAND gate

1602

outputs a signal DMPX

7

to a selector

1603

only when black is printed at a black edge in the full color standard copy mode. The selector

1603

selects the value data MVIDEO

7-0

subjected to the masking processing or the density data VIDEO

7-0

according as the signal DMPX

7

is L level or not.

The selector

1603

receives image data MVIDEO

7-0

subjected to masking processing at A input and image data VIDEO

7-0

converted to density at B input in the order or cyan, magenta, yellow and black. The data selected by the selector

1603

is supplied, through a line memory

1604

storing data of 5*5 matrix to a Laplacian filter, to a Laplacian filter

1605

, smoothing filters

1607

,

1608

and

1609

, a filter

1612

for detecting a minimum in a 5*5 matrix, a filter

1613

for detecting a minimum in a 3*3 matrix, and a print edge corrector

1615

.

The Laplacian filter

1605

, shown in

FIG. 31

, converts a data on a pixel under interest at the center to an enhanced data, and sends it to a DMTF table

1606

. The DMTF table performs conversion shown in FIG.

32

and sends a conversion data as density edge emphasis component data DMTF

7-0

.

The smoothing filters

1607

,

1608

and

1609

smoothens the input data to 300, 200 and 100 dpi, and

FIGS. 33-35

show examples of the three filters. The data subjected to smoothing as well as the data without subjected to smoothing is sent to a controller

1610

for smoothing filters. The controller

1610

also receives the change signal SH

2-0

from the HVC converter

114

. The controller

1610

selects one of the input data according to the change signal SH

2-0

and sends it as SD

7-0

. The change signal SH

2-0

is also received by another controller

1611

of edge emphasis coefficient to select one of eight kinds of the edge emphasis coefficients as ED

7-0

per each pixel (in real time), and change a plurality of sharpness up to eight areas simultaneously.

The filters

1612

and

1613

detect a minimum in a 5*5 matrix and in a 3*3 matrix if a pixel under interest is placed at the center of the matrices and they sent the results to a selector

1614

. The selector

1614

selects one of them according to a selection signal FSEL

2

, and it sends it as MIN

7-0

. The selection signal FSEL

2

has been determined experimentally. As explained above, by using the minimum data MIN

7-0

, image data can be decreased to zero only inside a black character, and the black character can be printed with edge emphases without white peripheral lines as shown in FIG.

37

A. On the other hand, if image data of cyan, magenta or yellow is subtracted by, for example, an edge detection quantity (such as FL

17-10

or FL

27-20

in this embodiment), undesired white peripheral lines as shown in

FIG. 37A

are observed.

The print edge corrector

1615

performs edge correction by taking into account a print characteristic on transferring a toner image onto a sheet of paper. The print characteristic means that more toners adhere to a start position while less toners adhere to an end position, as shown in

FIG. 38B

with a solid line. However, it is desirable that equal quantities of toners adhere to the start and end and positions. Such print characteristic occurs when image data changes largely at edges while a data near the edges is about zero. Then, the corrector

1615

corrects the data shown in

FIG. 38A

as shown in FIG.

38

B. Then, as shown in

FIG. 38B

with a dashed line, the inequality can be reduced.

FIG. 39

shows the print edge corrector

1615

in detail. If a data under interest is a data of an l-th pixel, a subtractor

1650

subtracts a data of (l+1)-th pixel from the data of the l-th pixel and sends the result to a comparator

1553

. If the result is larger than a threshold value REF

17-10

, the comparator

1653

sends a signal to input S

0

of a selector

1655

. A subtractor

1651

subtracts a data of the l-th pixel from the data of the (l−1)-th pixel and sends the result to a comparator

1554

. If the result is larger than a threshold value REF

27-20

, the comparator

1654

sends a signal to input S

1

of the selector

1655

. Further, if the data of the l-th data is smaller than a threshold value REF

37-30

, a comparator

1652

sends a signal to input S

2

of the selector

1655

.

If the selector

1655

receives L level at the input S

2

-S

0

, the pixel under interest is considered to exist between edges as shown in FIG.

40

B. In this case, the selector

1655

selects PD

7-0

after addition as ADD

17-10

. If the selector

1655

receives H level at the input S

1

and L level at the inputs S

0

and S

2

, the pixel under interest is considered to exist at a leading edge and below a reference level as shown in FIG.

40

A. In this case, the selector

1655

selects PD

17-10

as ADD

17-10

. Further, if the selector

1655

receives H level at the input S

0

and L level at the inputs S

1

-S

2

, the pixel under interest is considered to exist at a trailing edge and below a reference level as shown in FIG.

40

C. In this case, the selector

1655

selects PD

27-20

as ADD

17-10

.

Next, the MTF correction performed by the MTF corrector shown in

FIG. 29B

is explained. As explained previously, selectors

1616

and

1617

select one of value edge component VMTF

7-0

, density edge component DMTF

7-0

and edge emphasis quantity of zero according to the signals DMPX

0

and DMPX

1

on the kind of pixel DMPX

0

and DMPX

1

. The signals DMPX

0

and DMPX

1

are defined in Tables 3-8 in the various modes and output by the controller

1610

of the MTF correction parameters.

A selector

1622

receives ED

7-0

set by the CPU

1

directly and through multipliers

1619

-

1621

which multiply it with

¾, {fraction (1/2

+L )} and

{fraction (1/4+L )}, and selects one of the four inputs according to parameters DMPX3 and DMPX2. Another selector 1623 receives the output of the selector 1622 and the zero, and selects one of the two inputs according to a parameter DMPX4. As shown in Table

10, the parameters DMPX

4

-DMPX

2

are determined according to values of {overscore (AMI

3

)}-{overscore (AMI

0

)}. If all of {overscore (AMI

3

)}-{overscore (AMI

0

)} are H level or the pixel is not in a dot image, the edge emphasis coefficient ED

7-0

is sent readily as ED

17-10

to an operator

1618

. As explained previously, the region discriminator

146

changes {overscore (AMI

0

)}-{overscore (AMI

3

)} to L level successively as the degree of dot image increases. Then, the controller

1601

for the MTF correction parameters changes DMPX

4

-DMPX

1

according to the degree of dot image, and the selectors

1622

and

1623

suppress edge emphasis coefficients ED

7-0

according to results of dot detection {overscore (AMIO)}-{overscore (AMI

3

)}. The operator

1618

multiplies the edge emphasis quantity USM

7-0

with the edge emphasis coefficient ED

17-10

and divides the product with

128

to output USM

17-10

.

TABLE 10

Decision of dot image

{overscore (AMI3)}

{overscore (AMI2)}

{overscore (AMI1)}

{overscore (AMI0)}

DMPX4

DMPX3

DMPX2

ED

L

L

L

L

L

—

—

0

H

L

L

L

H

L

L

ED/4

H

H

L

L

H

L

H

ED/2

H

H

H

L

H

H

L

3ED/

4

H

H

H

H

H

H

H

ED

A selector

1626

receives data SD

7-0

directly and through a smoothing filter

1625

and selects one of the inputs according to DMPX

5

. Further, another selector

1627

selects one of the output of the selector

1627

and MIN

7-0

according to DMPX

6

to output VIDEO

17-10

. The control signals DMPX

5

and DMPX

6

are determined as shown in Tables 3-8.

An adder

1624

adds the edge emphasis quantity USM

17-10

to the pixel data VIDEO

27-20

. Another adder

1628

adds VIDEO

27-20

to ADD

17-10

to output as VIDE

0

37-30

. As explained above, the addition data ADD

17-10

is provided to add a pixel data at a leading edge or at a trailing edge.

I. Gamma Corrector

The gamma corrector

150

shown in

FIG. 41

receives the image data VIDEO

37-30

after the MTF correction, and it changes gamma correction curve according to an instruction by a user and corrects the image data to data of desired image quality. The image data VIDEO

37-30

and the change signal GA

2-0

for changing the gamma correction table are received by a gamma correction table

1702

. The change signal GA

2-0

are set by the HVC converter

114

. The table

1702

changes eight gradation curves shown in

FIGS. 42 and 43

in real time according to the change signal GA

2-0

as a BANK signal of the table.

FIG. 42

shows gradation curves in correspondence to the change signal GA

2-0

in the value control mode, while

FIG. 43

shows gradation curves in correspondence to the change signal GA

2-0

in the contrast control mode. The gamma correction table

1702

changes input data Din

7-0

(VIDEO

37-30

) to output data Dout

7-0

(VIDEO

47-40

).

An operator

1703

operates Eq. (20) based on the data VIDEO

47-40

output from the gamma correction table

1702

:

VIDEO

77-70

=(VIDEO

47-40

−UDC

7-0

)·GDC

7-0

/128, (20)

≦256.

That is VIDEO

77-70

=

256

if the operation at the left side exceeds

256

. As shown in Table 11, background clearance data UDC

7-0

and slope correction data GDC

7-0

have eight kinds of data.

TABLE 11

Background clearance data UDC and

slope correction data GDC

GDC

7-0

UDC

7-0

7

152

0

6

144

0

5

136

0

4

128

0

3

136

16

2

128

16

1

120

16

FIG. 44

shows a graph of VIDEO

77-70

plotted against VIDEO

47-40

for various values of CO

2-0

from 7 to 1. As shown in

FIG. 45

, background data UDC

7-0

is subtracted from VIDEO

47-40

and the slope is corrected by slope correction data GDC

7-0

.

Although the present invention has been fully described in connection with the preferred embodiments thereof with reference to the accompanying drawings, it is to be noted that various changes and modifications are apparent to those skilled in the art. Such changes and modifications are to be understood as included within the scope of the present invention as defined by the appended claims unless they depart therefrom.

Number	Name	Date
4978226	Moriya et al.	Dec 1990
5165071	Moriya et al.	Nov 1992
5181105	Udagawa et al.	Jan 1993
5208663	Hiratsuka et al.	May 1993
5408343	Sugiura et al.	Apr 1995
5675664	Maeda et al.	Oct 1997
5696842	Shirasawa et al.	Dec 1997
5699492	Karaki	Dec 1997
5726780	Hirota et al.	Mar 1998
5742410	Suzuki	Apr 1998

Color image processor

Information

Patent Number

Date Filed

Date Issued

Inventors

Original Assignees

Examiners

Agents

CPC

US Classifications

Field of Search

US

International Classifications

Abstract

Description

Claims

Priority Claims (1)

US Referenced Citations (10)

Non-Patent Literature Citations (2)

Entry
U.S. application No. 08/611,754, Co-pending.
U.S. application No. 08/747,827, Co-pending.