This application is based upon and claims the benefit of priority from the prior Japanese Patent Application No. 2006-123756, filed on Apr. 27, 2006, the entire contents of which are incorporated herein by reference.
1. Field of the Invention
The present invention relates to a detection device and detection method for detecting a repeat field in a telecine image transformed into 60 fields by 32-pull down, and more particularly to a detection device or the like for detecting a 32-pull down sequence at high reliability.
2. Description of the Related Art
When an image (cinema image) based on a 24 frame progressive scan is transformed into an image (telecine image) based on a 60 field interlace scan, a transformed processing called a 32-pull down has been conventionally used.
The 32-pull down transforms the cinema image into the telecine image by repeating a field two fields before once every five fields (e.g. Japanese Patent Application Laid-Open No. 2002-354431).
By this, the cinema image recorded on a DVD (Digital Versatile Disk) can be transferred to a display device as NTSC type TV signals, and the cinema image can be transmitted from a broadcast station as NTSC type TV signals.
In the case of displaying the 60 field telecine image transferred from the DVD, or transmitted from the broadcast station on the 60 frame progressive type display device, the 32-pull down sequence is detected (particularly the field repeated by 32-pull down sequence (hereafter called “repeat field”) is detected), and the interlace image is transformed into the progressive image (IP (Interlace to Progressive) transform).
In the IP transform, the 60 fields telecine image is transformed into the original 24 frame progressive image, by deleting the detected repeat fields, for example. And the telecine image is transformed into the progressive image by repeating two frames and repeating three frames for every other frame.
The DVD player 100 has a decoder 101 for decoding the telecine image encoded by compression, for example, and a converter 102 for IP-transforming the decoded telecine image. The repeat field is detected in the converter 102.
The broadcast station 120 has a converter 121 for performing 32-pull down transform, and an encoder 122 for encoding, such as compression.
By the converter 121, the cinema image is transformed into the telecine image, and is output to a display device 150 in a home as an analog broadcast. The telecine image is encoded by compressing, for example, by the encoder 122, and is output to the display device 140 via an antenna 130 as a digital broadcast.
IP transform is performed by the decoder 131 on the display device 140 in the case of the digital broadcast, and is performed by the display device 150 in the case of the analog broadcast.
In IP transform, it is critical to detect repeat fields. Conventionally the difference value of each pixel between adjacent fields is determined for the telecine image which is input, and this difference value is compared with a redundant threshold and a non-redundant threshold (e.g. Japanese Patent Application Laid-Open No. 2002-354431).
In another prior art, the repeat field is detected by comparing a cumulatively added difference value with a threshold value for each field period (e.g. Japanese Patent Application Laid-Open No. H10-65964).
However in the case of detecting repeat fields according to Japanese Patent Application Laid-Open No. 2002-354431 and No. H10-65964, if the telecine image broadcasted from the broadcast station is IP-transformed, for example, the progressive image cannot be displayed at high image quality because of the influence of noise.
In the example in
Also the characteristics of noise greatly differs between the case of performing IP transform by the DVD player (see
With the foregoing in view, it is an object of the present invention to provide a detection device and detection method for a 32-pull down sequence which can detect repeat fields, at high precision, from an interlace type telecine image transformed by 32-pull down.
To achieve the above object, the present invention is a detection device having a difference distribution information generation unit which generates difference distribution information between frames for an interlace image which is input, and an information table creation unit which creates n types (n is an integer) of information tables from the difference distribution information and n types of judgment conditions, wherein a 32-pull down sequence is detected from m fields (m is 5 or higher integer) of information in the past in the information table.
The detection device of the present invention further has a cumulative counter which accumulates one field of the generated difference distribution information, wherein the information table creation unit creates the information tables from the judgment conditions and one field of the difference distribution information accumulated by the cumulative counter.
The detection device of the present invention further has a judgment unit which judges a current field as a repeat field based on the information tables when information to indicate the repeat field is stored in the information tables for the current field and a field which is five fields before, and information to indicate an another field other than the repeat field is stored in the information tables for fields therebetween.
In the detection device of the present invention, the information table creation unit judges whether each of the fields has a noise or not, based only on the difference distribution information.
In the detection device of the present invention, the judgment unit judges that a current field is a still image when the information to indicate the repeat field is continuously stored in the information tables.
The detection device of the present invention further has a motion compensation control unit which control motion compensation for the input interlace image when the current field is judged as the still image by the judgment unit.
In the detection device of the present invention, the n is “5”, and the information table creation unit creates the information tables by comparing the difference distribution information with five types of threshold values.
In the detection device of the present invention, information of a current field and the past 10 fields, a total of 11 fields, is stored in the information tables.
To achieve the above object, the present information is a detection method having the steps of: generating difference distribution information between frames for an interlace image which is input; creating n types (n is an integer) information tables from the difference distribution information and n types of judgment conditions; and detecting a 32-pull down sequence from m fields (m is 5 or higher integer) of information in the past in the information tables.
According to the present invention, a detection device and detection method for a 32-pull down sequence can detect repeat fields, at high precision, from an interlace type telecine image transformed by 32-pull down.
Embodiments of the present invention will now be described with reference to the drawings.
The three field memories 11 to 13 store three fields of input images (F (t), F (t−1), F (t−2): t is time) which are continuous in time respectively. The input image is 60 fields of an interlace image (telecine image including repeat fields) after 32-pull down.
The pixel difference absolute value subtractor 20 reads a current fields image (F (t)) and field image two fields before (F (t−2)) from the field memories 11 and 13, subtracts the grayscale value at a same pixel position, and extracts the difference absolute value (frame difference).
The extraction of the difference absolute value is because of detecting the distribution of the frame difference between odd fields (top fields) or between even fields (bottom fields) of each pixel in field units. Frame difference distribution information is generated by the pixel difference absolute value subtractor 20.
If a pixel position is expressed by (x, y), the pixel difference absolute value subtractor 20 computes the following formula.
Difference absolute value(diff)=abs(F(t,x,y)−F(t−2,x,y) (1)
The absolute value difference comparator 30 compares the extracted difference absolute value and five threshold values, and outputs the comparison result.
The first threshold value (S11_TH) is a threshold value for checking whether the difference absolute value exists within four grayscales. This is the threshold value to be used for judging whether the input image includes noise, and for judging the repeat field. The value of the first threshold value (S11_TH) is “5”, for example.
The second threshold value (S12_TH) is a threshold value for checking whether the difference absolute value does not exist in 16 grayscales or later. This is the threshold value to be used for judging whether it is the repeat field when the input image does not have noise. The value of the second threshold value (S12_TH) is “15”, for example.
The third threshold value (S21_TH) is a threshold value for checking whether the difference absolute value does not exist after 32 grayscales or later. This is the threshold value to be used for judging whether it is another field other than the repeat field when the input image does not have noise. The value of the third threshold value (S21_TH) is “31”, for example.
The fourth threshold value (S22_TH) is a threshold value for checking whether the difference absolute value does not exist in 60 grayscales or later. This is the threshold value to be used for judging whether it is the repeat field when the input image has noise. The value of the fourth threshold value (S22_TH) is “59”, for example.
The fifth threshold value (S31_TH) is a threshold value for checking whether the difference absolute value does not exist in 80 grayscales or later. This is the threshold value to be used for judging whether it is another field other than the repeat field when the input image has noise.
The value of the fifth threshold value (S31_TH) is “79”, for example.
The absolute value difference comparator 30 compares these five threshold values and the difference absolute value respectively. In other words, if the difference absolute value is y−diff, then it is checked whether a pixel that satisfies
S11_TH>y_diff (2)
exists for the first threshold value (S11_TH) by comparison. For the second threshold value (S12_TH) to the fifth threshold value (31_TH), it is checked whether a pixel which satisfies
S12_TH<y_diff (3)
S21_TH<y_diff (4)
S22_TH<y_diff (5)
S31_TH<y_diff (6)
exists by comparison respectively.
The absolute value difference comparator 30 outputs “1” respectively if pixels satisfy Formulas (2) to (6). In the present embodiment, the absolute value difference comparator 30 is comprised of a total of five absolute value difference comparators 30 so as to compare each threshold value and the difference absolute value by Formulas (2) to (6).
The cumulative adder 40 cumulatively adds the output result of the absolute value difference comparator 30 for one field. In other words, the cumulative adder 40 counts the number of pixels that satisfy Formulas (2) to (6) for the pixels in one field, and outputs the count values (S11COUNT, S12COUNT, S21COUNT, S22COUNT, and S31COUNT respectively).
In the present embodiment, the cumulative adder 40 is comprised of five cumulative adders 40 corresponding to the absolute value difference comparators 30 in the previous stage.
The flag information creation processing unit 50 creates a flag information table indicating whether each field of the input image (more specifically, the current field F(t)) is the repeat field without noise or another field without noise or the repeat field with noise, or another field with noise based on each count value (e.g. S11COUNT).
If the field of the input image is judged as the repeat field without noise, the flag information creation processing unit 50 sets (or writes) [1] to the item [Field] [0] of the flag information table, and sets “0” to the other items. In the same way, “1” is set in each applicable item. These are judged for each field.
Specifically, this judgment is performed as follows. First a difference value ratio (Sum) is calculated from the count value within the four grayscale difference (S11COUNT). For the calculation formula,
Difference value ratio(Sum)=(S11COUNT/(total number of pixels in one field))*100 (7)
is used. This difference value ratio Sum indicates the ratio of pixels within “4” grayscales in each frame.
And the field is judged as the repeat field without noise when the difference value ratio (Sum) based on Formula (7) is “99%” or more, and the counter value (S12COUNT) is “0” (the difference absolute value does not exist in 16 grayscales or later, that is, the frame difference does not exist in 16 grayscales or later), and “1” is set for the [Field] [0] (which indicates the repeat field without noise) of the flag information table.
In this way, by the combination of the difference value ratio Sum (frame difference ratio) and the count value (e.g. S11COUNT), it is judged whether each field of the input image is the repeat field without noise, or the repeat field with noise, and so on. The reason for this is as follows.
In other words, in the case of such the field, almost all the pixels exist within the frame difference “15”. And the ratio of the pixels of which frame difference is within “4” grayscales is “99%” or more in the statistics, as shown in
This means that if the ratio of pixels of which frame difference is within “4” grayscales (difference value ratio Sum) is “99%” or more and if there are no pixels of which frame difference value is “16” grayscales or more, then it can be judged that the current field F(t) does not have noise, and is the repeat field.
In other words, when the difference absolute value Sum in Formula (7) is “99%” or more, and the count value S12COUNT that satisfies Formula (3) is “0”, then it can be judged that this field does not have noise, and is the repeat field.
The above mentioned second threshold (S11_TH) is set to “15”, so as to correspond to the frame difference value “15” shown in
In the same way, in area “B” of
Therefore if the ratio of the pixels of which frame reference is within “4” grayscales (difference value ratio Sum) is “90%” or less, or if pixels exist in an area of which the frame difference value is “16” grayscales or later, then it can be judged that the current field F(t) is the field without noise, and is the another field.
In the present embodiment, to judge whether the current field F(t) is the another field without noise with certainty, it is judged whether pixels do not exist in an area where the frame difference value is “32” grayscales or later.
In other words, if the difference value ratio (Sum) of Formula (7) is “90%” or less, or if the count value S21COUNT that satisfies Formula (4) is not “0”, then it can be judged that the current field F(t) is the field without noise, and is the another field other than the repeat field. The third threshold value (S21_TH) is set to “31” for this reason.
Therefore, it can be judged that the current field F(t) is the field with noise, and is the repeat field when the difference value ratio (Sum) of Formula (7) is “70%” or more, and the count value S22COUNT that satisfies Formula (5) is “0”. The fourth threshold (S22_TH) is set to “59” for this reason.
In area “D”, each pixel of the current field F(t), which is the field with noise and is the another field other than the repeat field, exists. Statistically, the ratio of pixels of which the frame difference is within “4” grayscales is “60%” or less.
Therefore if pixels exist in an area of which difference value ratio (Sum) is “60%” or less, or the frame difference value is “60” grayscales or later, it can be judged that the current field F(t) is the another field with noise.
In the present embodiment, to judge whether the current field F(t) is the another field with noise with certainty, it is judged whether pixels do not exist in an area where the frame difference value is “80” grayscales or later.
Therefore if the difference value ratio (Sum) of Formula (7) is “60%” or less, or if the count value S31COUNT that satisfies Formula (6) is not “0”, it can be judged that the current field F(t) is the another field with noise.
In this way, according to the present embodiment, it is judged whether the current field F(t) has noise or not based on the frame difference ratio (difference value ratio (Sum)) in order to create the flag information table, and it is also judged whether the current field is the repeat field or another field by changing the frame difference value (that is, the first to fifth threshold values) depending on whether noise exists or not.
The presence of noise is judged by the frame difference ratio because the frame difference when noise is absent is confined in a range from “80%” to “90%” statistically (or as a result of experiment), as mentioned above. Each value to indicate the difference value ratio (Sum) is set to “90%” or “70%”, so as to judge the presence of noise with certainty.
Each value of the difference value ratio (Sum) is an example, and may be other values. Therefore it can be judged that the current field is the another field without noise when the difference value ratio (Sum) is “98%” or less or the count value S21COUNT is not “0”, and that the current field is the repeat field with noise when the difference value ratio (Sum) is “65%” or more, and the count value S22COUNT is not “0”, and that the current field is the another field with noise when the difference value ratio (Sum) is “64%” or less or the count value S31COUNT is not “0”.
Based on such judgment, the flag information creation processing unit 50 stores each value for each field in each item of the flag information table. In the present embodiment, two flag information tables, one for the case of the no noise state and another for the case of the noise state, are provided.
The repeat field judgment unit 60 judges whether the current field is the repeat field or not based on the created flag information table. This repeat field judgment unit 60 actually performs judgment on whether the current field is the repeat field without noise or not and performs other judgments mentioned above.
In the present embodiment, information on the current field and past 10 fields, a total of 11 fields, are stored in the flag information table.
This is because the repeat field is repeated once every five fields in the input image by 32-pull down, and whether the current field is the repeat field or not can be detected only when this sequence is detected. If the repeat field is detected only in the past five fields, and then 60 frames of the progressive image is displayed by IP transform, the image is displayed like a “comb”, and the image deteriorates. Therefore, in order to prevent a detection error of the 32-pull down sequence and to prevent image deterioration, a margin is provided to the sequence, and information on the past 10 fields is stored.
Therefore the repeat field judgment unit 60 judges that the current field is the repeat field by checking that the 32-pull down sequence, in which “0” or “1” is repeated in the past four fields and the fifth field in the past is “1” or “0”, is repeated twice in the information table.
Based on the judgment result of the repeat field judgment unit 60, the 32-transform control unit 70 outputs 60 fields of the progressive image. Actual IP transform is performed by the 32-transform control unit 70.
By judging the repeat field, the 32-transform control unit 70 regards the field as a pair of repeat field, not as a field in the back but as a field in the front on the time base, and combines this repeat field and the front field, and outputs them as one frame. By this combination, an output image, of which image quality is equivalent to the input image, without image deterioration, can be acquired. And 60 fields of the progressive image with no repeat filed can be output. Clearly other combinations are possible if the output image is not deteriorated.
The statistical information extraction processing is a processing for computing the frame difference value (difference absolute value (diff)) in the pixel difference absolute value subtractor 20 (computing Formula (1)), comparing the result with each threshold value in the absolute value difference comparator 30 (computing Formula (2) to Formula (6)), and acquiring one field of the count value (e.g. S11COUNT) in the cumulative adder 40.
Then in the flag information creation processing 50, the flag information table creation processing is executed (S30).
When this creation processing is started (S31), the statistical information (count value) is read first (S32). The flag information creation processing unit 50 reads each counter value (e.g. S11COUNT) accumulated by the cumulative adder 40.
And Formula (7) is computed from the count value S11COUNT, and the frame difference ratio (Sum) is determined, and it is judged whether the frame difference ratio (Sum) is “99%” or more and the count value S12COUNT is “0” or not (S33). If this condition is satisfied (YES), “1” is stored in the item [Field] [0] of the flag information table (S34). If not (NO in S33), “0” is stored in the item [Field] [0] (S36). Then this creation processing (S35 in
On the other hand, it is also judged whether the frame difference ratio (Sum) is “90%” or less or the count value S21COUNT is not “0” (S37). If this condition is satisfied (YES), “1” is set in the item [Field] [1] of the flag information table (S38), and if not (NO in S37), “0” is set (S39). And this creation processing ends (S35).
It is also judged whether the frame difference ratio (Sum) is “70%” or more, and the count value S22COUNT is “0” (S40 in
Also it is judged whether the frame difference ratio (Sum) is “60%” or less or the count value S31COUNT is not “0” (S43). If this condition is satisfied (YES), “1” is set in the item [Field] [3] of the flag information table (S44), and if not (NO in S43), “0” is set (S45). And this creation processing ends (S35).
Back in
When this detection processing starts (S51), the flag information table is read (S52). And it is detected whether the current field has noise or not from the flag information table (S53).
It is judged that the current field is the field without noise if the flag “1” is stored in the item [Field] [0] or the item [Field] [1] of the flag information table, and it is judged that the current field is the field with noise if the flag “1” is stored in the item [Field] [2] or item [Field] [3].
When noise exists (YES), information on the past N fields is referred to in the information table for the no noise state (S55). This is for detecting the 32-pull down sequence.
And it is judged whether the flag to indicate the repeat field exists once every five fields, and there are four flags to indicate another field continuously there between (S55). If this condition is satisfied (YES), it is judged that the current field is the repeat field without noise (S56). And this detection processing ends (S57).
If this condition is not satisfied (NO in S55), it is judged that the current field is the another field without noise (S58). And this detection processing ends (S57).
When the current field has noise (NO in S53), past N fields are referred to in the flag information table for the noise state (S59). And it is judged whether the flag to indicate the repeat field exists once every five fields, and there are four flags to indicate another field continuously there between (S60).
If this condition is satisfied (YES), it is judged that the current field is the repeat field with noise (S61). If not (NO in S60), it is judged that the current field is the another field with noise (S62). And this detection processing ends (S57).
In this detection processing, there are two information tables, one for the case of the noise state and another for the case of the no noise state, so the presence of noise is judged in S53, then the same processing is executed respectively.
Back in
And the entire processing ends (S80).
In this way, the flag information table is created from the difference value ratio (Sum) and the count value (e.g. S12COUNT), the past N fields are referred to in this table, and the 32-pull down sequence is detected first, then it is detected whether the current field is the repeat field or not, so repeat fields can be detected at high precision.
Also 60 frames of the progressive image is generated by 32-transform control based on the detected repeat fields, so the output image equivalent to the input image without image deterioration can be acquired.
With the above example, if the input image is a still image, “1” or “0”, the same as the repeat field without noise, is stored in the item [Field] [0] and [Field] [1] of the flag information table continuously for several fields.
In the present embodiment, if “1” or “0” is stored continuously for “5” fields, the repeat field judgment unit 60 judges that the continuous fields are fields of the still image. For certain it may be judged as the fields of the still image when a number of fields, other than “5” are stored.
If 32-transform control is performed for the field judged as the still image like this, normally the sixth field is damaged, and the output image quality deteriorates.
Therefore, as shown in
In the motion compensation, a motion vector is detected in the image, and the estimated error is corrected based on this information to acquire the correct image. Such motion compensation can be implemented by a known configuration and operation.
In the example in
In the above mentioned example, the specific value of the first threshold value (S11_TH) was described as “5”. Needless to say, this is an example, and any threshold value, if the value satisfies the frame difference distribution shown in
Also in the above example, the grayscale value of the input image was described as “8” bits (“256” grayscales). The present invention can be applied to the input image having another number of bits (grayscales), and the same effect can be exhibited.
The present detection device 1 can be suitably applied to such information equipment as a personal computer, such a display device as a liquid crystal TV, and other equipment that can display progressive type images.
Number | Date | Country | Kind |
---|---|---|---|
2006-123756 | Apr 2006 | JP | national |