Alpha blending calculator

BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates to an alpha blending calculator which executes an alpha blending calculation .alpha.X+(1-.alpha.)Y in accordance with given digital data X, Y, and .alpha..
2. Description of the Related Art
In recent years, as digital techniques progress, images are more frequently processed in the form of digital data. In various apparatuses handling such digital images, a process called "alpha blending" is often used. The various apparatuses include, for example, an apparatus for applying special effects to images, an apparatus for processing images, and an apparatus for producing images. In the alpha blending process, two images are blended at a predetermined ratio. The alpha blending is accomplished by executing an operation of Expression (1) below for every pixel included in two images. Hereinafter, the operation shown in Expression (1) is referred to as "an alpha blending calculation".
P=.alpha.X+(1-.alpha.)Y (1)
where X denotes a digital value of a pixel included in one image, Y denotes a digital value of a pixel included in another image, and .alpha. denotes a predetermined blending ratio.
In order to obtain a single blended image by blending two images, it is necessary to execute an enormous number of alpha blending calculations. This means that it is necessary to execute each alpha blending calculation at a high speed.
Japanese Laid-Open Patent Publication No. 6-83852 discloses a conventional alpha blending calculator. Hereinafter, the conventional alpha blending calculator will be described with reference to FIGS. 7 and 8.
FIG. 7 shows the organization of an alpha blending calculator which executes an alpha blending calculation .alpha.X+(1-.alpha.)Y in accordance with given digital data X, Y, and .alpha.. Each of the digital data X, Y, and .alpha. is 4-bit binary data. In FIG. 7, X.sub.0 through X.sub.3 represent the least significant bit (LSB) to the most significant bit (MSB) of the digital data X, respectively. Also, Y.sub.0 through Y.sub.3 represent the least significant bit (LSB) to the most significant bit (MSB) of the digital data Y, respectively. The digital data .alpha. is a fixed-point number with the binary point to the left of the most significant bit (MSB). The least significant bit (LSB) to the most significant bit (MSB) of the digital data .alpha. are represented by .alpha..sub.0 through .alpha..sub.3, respectively. The reference symbols P.sub.0 through P.sub.7 represent the least significant bit (LSB) to the most significant bit (MSB) of an 8-bit fixed-point number P indicative of the result of the alpha blending calculation .alpha.X+(1- .alpha.)Y. The binary point is located between P.sub.3 and P.sub.4.
The alpha blending calculator includes multiplexers 71 arranged in 4 rows and 4 columns, and an adding section 74 for adding the outputs from the multiplexers 71. The adding section 74 includes half adders 72, full adders 73, and an adder 75 which executes multi-bit addition.
Each of the multiplexers 71 has two data lines for receiving two data inputs and one selection line for receiving selection data. The multiplexer 71 outputs one of two data inputs in accordance with the selection data. Respective bits of the digital data X and Y are input through the data lines of the multiplexers 71, and respective bits of the digital data .alpha. are input through the selection lines of the multiplexers 71.
The outputs of the multiplexers 71 are connected to the half adders 72 and the full adders 73. The half adders 72 and the full adders 73 are arranged in accordance with the organization of array multiplier.
As described above, in the case where the digital data .alpha. is composed of 4 bits, Expression (1) can be transformed into Expression (2) below.
P=.alpha.X+.alpha.Y+Y2.sup.-4 ( 2)
When Expression (2) is expanded while the third term of the right side is omitted, Expression (3) below can be obtained. ##EQU1##
Therefore, Expression (3) is the approximation of Expression (2). However, if the number of bits of the digital data .alpha. is increased, the difference between P obtained by Expression (2) and P obtained by Expression (3) becomes small and negligible.
In the alpha blending calculator shown in FIG. 7, the outputs of the multiplexers 71 in the first row and the outputs of the multiplexers 71 in the second row are added. The results of the addition and the outputs of the multiplexers 71 in the third row are added. The results of the addition and the outputs of the multiplexers 71 in the fourth row are added. Then the results of the addition are added considering the carry. As a result, P is calculated on the basis of Expression (3).
FIG. 8 shows the organization of another conventional alpha blending calculator for obtaining complete P shown in Expression (2). In FIG. 8, the meanings of X.sub.0 -X.sub.3, Y.sub.0 -Y.sub.3, .alpha..sub.0 -.alpha..sub.3, and P.sub.0 -P.sub.7 are the same as those described above, and the descriptions thereof are therefore omitted. The alpha blending calculator shown in FIG. 8 includes multiplexers 81 and an adding section 84 for adding the outputs from the multiplexers 81. The adding section 84 includes half adders 82, full adders 83, and an adder 85 which executes multi-bit addition.
The organization of the alpha blending calculator shown in FIG. 8 is different from the organization of the alpha blending calculator shown in FIG. 7 in that Y.sub.0 -Y.sub.3 are input to the adders in the first row (stage). Accordingly, the complete alpha blending calculation shown in Expression (2) can be executed.
The most easiest way for implementing an alpha blending calculator in accordance with Expression (1) is to combine one complementer, two multipliers and one adder. However, such an implementation disadvantageously results in an increase of a circuit scale of the alpha blending calculator. On the other hand, the alpha blending calculators shown in FIGS. 7 and 8 can be implemented in a circuit scale of about one multiplier.
The alpha blending calculators shown in FIGS. 7 and 8, however, utilize the organization of array multiplier to obtain the sum of the outputs of the multiplexers, so that a problem exists in that it is difficult to execute the calculation at a high speed.
SUMMARY OF THE INVENTION
The alpha blending calculator of this invention for executing an alpha blending calculation in accordance with digital data X, Y, and .alpha. each composed of a plurality of bits, includes: a plurality of selection means, each of the selection means selecting one of X or Y in accordance with one of the plurality of bits of the digital data .alpha.; and calculation means for shifting outputs of the plurality of selection means by a predetermined number of bits, respectively, and for calculating a sum of the shifted outputs, wherein the calculation means includes first level to Nth level first adding means, and adding means connected to outputs of the Nth level first adding means for executing a multi-bit addition, each of the first level to the Nth level first adding means receives a plurality of stages of data, classifies the plurality of stages of data into a plurality of quotient groups each including a predetermined number of stages of data and a remainder group including data of stages the number of which is smaller than the predetermined number, calculates a sum of the predetermined number of stages of data and a carry for each of the plurality of quotient groups, and outputs the sum, the carry, and the data of stages the number of which is smaller than the predetermined number classified into the remainder group to the next level first adding means, and the outputs of the plurality of selection means are input as the plurality of stages of data to the first level first adding means, and the outputs of the Nth level first adding means consist of a sum and a carry.
In one embodiment of the invention, each of the plurality of selection means is a multiplexer.
In another embodiment of the invention, each of the first level to Nth level first adding means includes a half adder and a full adder.
In another embodiment of the invention, the digital data Y is further input as the plurality of stages of data to the first level first adding means of the calculation means.
Thus, the invention described herein makes possible the advantage of providing an alpha blending calculator of a small circuit scale which is capable of executing an alpha blending calculation at a high speed.
This and other advantages of the present invention will become apparent to those skilled in the art upon reading and understanding the following detailed description with reference to the accompanying figures.

BRIEF DESCRIPTION OF THE DRAWINGS
FIG. 1 shows the process of an alpha blending calculation according to the invention.
FIG. 2 is a diagram showing the organization of an alpha blending calculator in one example according to the invention.
FIG. 3 shows another process of an alpha blending calculation according to the invention.
FIG. 4 is a diagram schematically showing the organization of an alpha blending calculator in another example according to the invention.
FIG. 5 shows another process of an alpha blending calculation according to the invention.
FIG. 6 is a diagram schematically showing the organization of an alpha blending calculator in another example according to the invention.
FIG. 7 shows the organization of a conventional alpha blending calculator.
FIG. 8 shows the organization of another conventional alpha blending calculator.

DESCRIPTION OF THE PREFERRED EMBODIMENTS
FIG. 1 shows the process of an alpha blending calculation according to the invention. In this example, each of digital data X, Y, and .alpha. is 4-bit binary data. In FIG. 1, X.sub.0 through X.sub.3 represent the least significant bit (LSB) to the most significant bit (MSB) of the digital data X, respectively. Also, Y.sub.0 through Y.sub.3 represent the least significant bit (LSB) to the most significant bit (MSB) of the digital data Y, respectively. The digital data .alpha. is a fixed-point number with the binary point to the left of the most significant bit (MSB). The least significant bit (LSB) to the most significant bit (MSB) of the digital data .alpha. are represented by .alpha..sub.0 through .alpha..sub.3, respectively. The reference symbols P.sub.0 through P.sub.7 represent the least significant bit (LSB) to the most significant bit (MSB) of an 8-bit fixed-point number P indicative of the result of the alpha blending calculation .alpha.X+(1-.alpha.)Y. The binary point is located between P.sub.3 and P.sub.4.
Hereinafter, with reference to FIG. 1, the procedure of the alpha blending calculation (step 1 to step 3) will be described.
In step 1, either one of the digital data X or Y is selected in accordance with each bit of the digital data .alpha.. The reference symbols Sel.sub.0, Sel.sub.1, Sel.sub.2, and Sel.sub.3 indicate the bit arrangements in which the data X or Y which are selected for the respective bits (.alpha..sub.3, .alpha..sub.2, .alpha..sub.1, .alpha..sub.0) of the digital data .alpha. are successively shifted to the left by one bit. A circle shows a bit of the selected data X or Y. The data X is selected when the associated bit of the data .alpha. is "1", and the data Y is selected when the associated bit of the data .alpha. is "0". In the case where (.alpha..sub.3, .alpha..sub.2, .alpha..sub.1, .alpha..sub.0)=(0, 1, 1, 0), Y, X, X, and Y are selected. As a result, in the first stage to the fourth stage, (Y.sub.3, Y.sub.2, Y.sub.1, Y.sub. 0), (X.sub.3, X.sub.2, X.sub.1, X.sub.0), (X.sub.3, X.sub.2, X.sub.1, X.sub.0), and (Y.sub.3, Y.sub.2, Y.sub.1, Y.sub.0) are arranged in such a manner so as to be successively shifted to the left by one bit. Next, respective digits of the Sel.sub.0, Sel.sub.1, and Sel.sub.2 are added to each other, so as to obtain a sum S.sub.0 and a carry C.sub.0. For example, the addition can be accomplished by adding the digits in blocks 10 by two half adders, and by adding the digits in blocks 11 by two full adders, as shown in FIG. 1.
In step 2, respective digits of the sum S.sub.0, the carry C.sub.0, and the Sel.sub.3 obtained in step 1 are added to each other, so as to obtain a sum S.sub.1 and a carry C.sub.1. For example, the addition can be accomplished by adding the digits in a block 10 by a half adder, and by adding the digits in blocks 11 by three full adders, as shown in FIG. 1.
In step 3, the sum S.sub.1 and the carry C.sub.1 obtained in step 2 are added considering the carry. As a result, an 8-bit fixed-point number P is obtained. The fixed-point number P is obtained in accordance with Expression (3) shown above. Accordingly, the fixed-point number P is the approximation of the alpha blending calculation .alpha.X+(1-.alpha.)Y.
FIG. 2 shows the organization of an alpha blending calculator in one example according to the present invention. The alpha blending calculator executes the alpha blending calculation in accordance with the procedure shown in FIG. 1. The alpha blending calculator includes multiplexers 21 arranged in four rows and four columns, and an adding section 22. Each of the multiplexers 21 selects X or Y in accordance with each bit of the digital data .alpha.. In the adding section 22, the outputs of the multiplexers 21 are shifted by a predetermined number of bits, and the sum of the shifted outputs is obtained. The adding section 22 has a first level adding portion 23, a second level adding portion 24, and an adder 25. The first level adding portion 23 is connected to the outputs of the multiplexers 21 arranged in the first to third rows, so as to execute a process corresponding to step 1 in FIG. 1. The first level adding portion 23 is implemented by two half adders (HA's) and two full adders (FA's). The second level adding portion 24 is connected to the outputs of the first level adding portion 23 and the outputs of the multiplexers 21 arranged in the fourth rows, so as to execute a process corresponding to step 2 shown in FIG. 1. The second level adding portion 24 is implemented by a single half adder (HA) and three full adders (FA's). The adder 25 is connected to the outputs of the second level adding portion 24, so as to execute a process corresponding to step 3 shown in FIG. 1.
A half adder (HA) receives two single bit numbers A and B, and outputs a sum S of A and B, and a carry C to the next digit. Each of A, B, S, and C is "0" or "1". A full adder (FA) takes into account a carry C.sub.in from the previous digit. The full adder receives two single bit numbers A and B, and the carry C.sub.in, and outputs the sum S of A and B, and a carry C.sub.out to the next digit. Each of A, B, S, C.sub.in, and C.sub.out is "0" or "1". In FIG. 2, the carry output from a half adder (HA) or a full adder (FA) is indicated by an arrow pointing to the lower left direction, and the sum output from a half adder (HA) or a full adder (FA) is indicated by an arrow pointing to the downward direction.
In the alpha blending calculator shown in FIG. 2, the number of levels of addition required before the adder 25 is 2. In the conventional alpha blending calculator having the organization of array multiplier shown in FIG. 7, the number of levels of the same addition is 3. Thus, according to the alpha blending calculator in this example, the number of levels of addition can be reduced, so that the process time can be shortened. In addition, the conventional alpha blending calculator having the organization of array multiplier shown in FIG. 7 includes three half adders and six full adders. On the other hand, the alpha blending calculator shown in FIG. 2 includes three half adders and five full adders. Thus, according to the alpha blending calculator in this example, the number of adders can be reduced, so that the circuit scale of the alpha blending calculator can be minimized.
FIG. 3 shows a process of an alpha blending calculation in another example according to the invention. In this example, it is assumed that each of X, Y, and .alpha. is 8-bit binary data. In FIG. 3, X.sub.0 through X.sub.7 represent the least significant bit (LSB) to the most significant bit (MSB) of the digital data X, respectively. Also, Y.sub.0 through Y.sub.7 represent the least significant bit (LSB) to the most significant bit (MSB) of the digital data Y, respectively. The digital data .alpha. is a fixed-point number with the binary point to the left of the most significant bit (MSB). The least significant bit (LSB) to the most significant bit (MSB) of the digital data .alpha. are represented by .alpha..sub.0 through .alpha..sub.7, respectively. The reference symbols P.sub.0 through P.sub.15 represent the least significant bit (LSB) to the most significant bit (MSB) of a 16-bit fixed-point number P indicative of the result of the alpha blending calculation .alpha.X+(1-.alpha.)Y. The binary point is located between P.sub.7 and P.sub.8.
Hereinafter, the procedure of the alpha blending calculation (step 1 to step 5) shown in FIG. 3 will be described.
In step 1, one of the data X or Y is selected in accordance with each bit of the digital data .alpha.. The reference symbols Sel.sub.0 to Sel.sub.7 indicate the bit arrangements in which the data X or Y selected for the respective bits (.alpha..sub.7, .alpha..sub.6, .alpha..sub.5, .alpha..sub.4, .alpha..sub.3, .alpha..sub.2, .alpha..sub.1, .alpha..sub.0) are successively shifted to the left by one bit. A circle shows a bit of the selected data X or Y.
The eight stages of data Sel.sub.0 to Sel.sub.7 are grouped by regarding three stages of data as a unit. As a result, Sel.sub.0 to Sel.sub.7 are classified into two quotient groups g.sub.0 and g.sub.1, and one remainder group g.sub.rem. The quotient group g.sub.0 includes three stages of data Sel.sub.0 -Sel.sub.2. The quotient group g.sub.1 includes three stages of data Sel.sub.3 -Sel.sub.5. The remainder group g.sub.rem includes two stages of data Sel.sub.6 and Sel.sub.7.
Respective digits of Sel.sub.0 to Sel.sub.2 classified into the quotient group g.sub.0 are added to each other, so as to obtain a sum S.sub.0 and a carry C.sub.0. For example, the addition can be accomplished by adding the digits in blocks 30 by two half adders, and by adding the digits in blocks 31 by six full adders, as shown in FIG. 3. Respective digits of Sel.sub.3 to Sel.sub.5 classified into the quotient group g.sub.1 are added to each other, so as to obtain a sum S.sub.1 and a carry C.sub.1. For example, the addition can be accomplished by adding the digits in blocks 30 by two half adders, and by adding the digits in blocks 31 by six full adders, as shown in FIG. 3.
In step 2, two stages of data (the sum S.sub.0 and the carry C.sub.0), two stages of data (the sum S.sub.1 and the carry C.sub.1), and the two stages of data Sel.sub.6 and Sel.sub.7 which are classified into the remainder group g.sub.rem in step 1 are grouped by regarding three stages of data as a unit. As a result, total six stages of data are classified into two quotient groups g.sub.2 and g.sub.3. The quotient group g.sub.2 includes three stages of data S.sub.0, C.sub.0, and S.sub.1. The quotient group g.sub.3 includes three stages of data C.sub.1, Sel.sub.6, and Sel.sub.7. When 6 is divided by 3, the remainder is 0. For this reason, there is no data which is classified into the remainder group g.sub.rem.
Respective digits of S.sub.0, C.sub.0, and S.sub.1 classified into the quotient group g.sub.2 are added to each other, so as to obtain a sum S.sub.2 and a carry C.sub.2. For example, the addition can be accomplished by adding the digits in a block 30 by a half adder, and by adding the digits in blocks 31 by seven full adders, as shown in FIG. 3. Respective digits of C.sub.1, Sel.sub.6, and Sel.sub.7 classified into the quotient group g.sub.3 are added to each other, so as to obtain a sum S.sub.3 and a carry C.sub.3. For example, the addition can be accomplished by adding the digits in blocks 30 by two half adders, and by adding the digits in blocks 31 by six full adders, as shown in FIG. 3.
In step 3, two stages of data (the sum S.sub.2 and the carry C.sub.2) and two stages of data (the sum S.sub.3 and the carry C.sub.3) are grouped by regarding three stages of data as a unit. As a result, the total four stages of data are classified into one quotient group g.sub.4 and one remainder group g.sub.rem. The quotient group g.sub.4 includes the three stages of data S.sub.2, C.sub.2, and S.sub.3. The remainder group g.sub.rem include one stage of data C.sub.3.
Respective digits of S.sub.2, C.sub.2, and S.sub.3 classified into the quotient group g.sub.4 are added to each other, so as to obtain a sum S.sub.4 and a carry C.sub.4. For example, the addition can be accomplished by adding the digits in blocks 30 by four half adders, and by adding the digits in blocks 31 by six full adders, as shown in FIG. 3.
In step 4, two stages of data (the sum S.sub.4 and the carry C.sub.4) and one stage of data C.sub.3 classified into the remainder group g.sub.rem in step 3 are grouped by regarding three stages of data as a unit. As a result, the total three stages of data are classified into one quotient group g.sub.5. The quotient group g.sub.5 includes three stages of data S.sub.4, C.sub.4, and C.sub.3. When 3 is divided by 3, the remainder is 0. For this reason, there is no data which is classified into the remainder group g.sub.rem.
Respective digits of S.sub.4, C.sub.4, and C.sub.3 classified into the quotient group g.sub.5 are added to each other, so as to obtain a sum S.sub.5 and a carry C.sub.5. For example, the addition can be accomplished by adding the digits in blocks 30 by four half adders, and by adding the digits in blocks 31 by seven full adders, as shown in FIG. 3.
In the above-described steps 1 to 4, the classification is executed in such a manner that the data of stages which are adjacent to each other are classified into the same group. Alternatively, the classification may be executed in such a manner that the data of stages which are not adjacent to each other are classified into the same group.
In general, when the number of data stages in step n is S.sub.n, the number of data stages in step (n+1) is represented by Expression (4) below.
S.sub.n+1 =(S.sub.n /3).times.2+(S.sub.n %3) (4)
In Expression (4), the symbol "/" indicates a calculation for obtaining a quotient, and the symbol "%" indicates a calculation for obtaining a remainder.
According to Expression (4), S.sub.n+1 <S.sub.n. Therefore, any number of data stages in step 1 can be finally reduced to be 2 by repeatedly executing the process finite number of times.
In step 5, the sum S.sub.5 and the carry C.sub.5 obtained in step 4 are added considering the carry. As a result, a 16-bit fixed-point number P is obtained. The fixed-point number P is obtained in accordance with Expression (3) shown above. Accordingly, the fixed-point number P is the approximation of the alpha blending calculation .alpha.X+(1-.alpha.)Y.
FIG. 4 schematically shows the organization of an alpha blending calculator which executes the alpha blending calculation in accordance with the procedure shown in FIG. 3. The alpha blending calculator includes a selector 400 for selecting X or Y in accordance with each bit of the digital data .alpha.. The selector 400 outputs Sel.sub.0 -Sel.sub.7 indicative of the selection results.
The alpha blending calculator further includes first level to fourth level adding portions 401-404, and an adder 411. The outputs Sel.sub.0 -Sel.sub.7 of the selector 400 are input into the first level adding portion 401. The first level adding portion 401 executes a process corresponding to step 1 shown in FIG. 3. The first level adding portion 401 is implemented by adding portions 405 and 406 each having two half adders (HA's) and six full adders (FA's). The adding portion 405 executes the adding process related to the quotient group g.sub.0 in step 1 of FIG. 3, and outputs the sum S.sub.0 and the carry C.sub.0 to the second level adding portion 402. The adding portion 406 executes the adding process related to the quotient group g.sub.1 in step 1 of FIG. 3, and outputs the sum S.sub.1 and the carry C.sub.1 to the second level adding portion 402. The outputs Sel.sub.6 and Sel.sub.7 of the selector 400 are not necessarily input into the first level adding portion 401, because the outputs Sel.sub.6 and Sel.sub.7 are not processed in the first level adding portion 401.
The outputs of the adding portion 405 (the sum S.sub.0 and the carry C.sub.0), the outputs of the adding portion 406 (the sum S.sub.1 and the carry C.sub.1), and the outputs Sel.sub.6 and Sel.sub.7 of the selector 400 are input into the second level adding portion 402. The second level adding portion 402 executes a process corresponding to step 2 of FIG. 3. The second level adding portion 402 is implemented by an adding portion 407 having a half adder (HA) and seven full adders (FA's) and an adding portion 408 having two half adders (HA's) and six full adders (FA's). The adding portion 407 executes the adding process related to the quotient group g.sub.2 in step 2 of FIG. 3, and outputs the sum S.sub.2 and the carry C.sub.2 to the third level adding portion 403. The adding portion 408 executes the adding process related to the quotient group g.sub.3 in step 2 of FIG. 3, and outputs the sum S.sub.3 and the carry C.sub.3 to the third level adding portion 403.
The outputs of the adding portion 407 (the sum S.sub.2 and the carry C.sub.2) and the outputs of the adding portion 408 (the sum S.sub.3 and the carry C.sub.3) are input into the third level adding portion 403. The third level adding portion 403 executes a process corresponding to step 3 of FIG. 3. The third level adding portion 403 is implemented by an adding portion 409 having four half adders (HA's) and six full adders (FA's). The adding portion 409 executes the adding process related to the quotient group g.sub.4 in step 3 of FIG. 3, and output the sum S.sub.4 and the carry C.sub.4 to the fourth level adding portion 404. The carry C.sub.3 which is output from the adding portion 408 is not necessarily input into the third level adding portion 403, because the carry C.sub.3 is not processed in the third level adding portion 403.
The outputs of the adding portion 409 (the sum S.sub.4 and the carry C.sub.4) and the carry C.sub.3 output from the adding portion 408 are input into the fourth level adding portion 404. The fourth level adding portion 404 executes a process corresponding to step 4 of FIG. 3. The fourth level adding portion 404 is implemented by an adding portion 410 having four half adders (HA's) and seven full adders (FA's). The adding portion 410 executes the adding process related to the quotient group g.sub.5 in step 4 of FIG. 3, and outputs the sum S.sub.5 and the carry C.sub.5 to the adder 411.
The adder 411 adds the sum S.sub.5 and the carry C.sub.5 considering the carry, and outputs the result of addition as the 16-bit fixed-point number P.
In the alpha blending calculator shown in FIG. 4, the number of levels of addition before the adder 411 is 4. In the conventional alpha blending calculator having the organization of array multiplier, the number of levels of the same addition is 7 (=8-1). Thus, according to the alpha blending calculator in this example, the number of levels of addition can be greatly reduced, so that the process time can be shortened. A conventional alpha blending calculator having the organization of array multiplier includes seven (=8-1) half adders in the first stage and forty-two (=7.times.(7-1)) full adders in the remaining stage. On the other hand, the alpha blending calculator shown in FIG. 4 includes 15 half adders and 38 full adders. Thus, the circuit scale of the alpha blending calculator in this example is slightly larger than that of the conventional one.
In general, the number of bits of the digital data X and Y is m, the number of levels of addition required for the conventional alpha blending calculator having the organization of array multiplier is (m-1), and hence the process time is 0(m). On the other hand, the level number L of adding processes required for the alpha blending calculator according to the invention is approximately based on Expression (5) below, so that the process time is 0(log m).
L=5.67(log m-0.30) (5)
Therefore, the alpha blending calculator according to the invention can execute the calculation at a much higher speed as compared with the conventional alpha blending calculator having the organization of array multiplier. It is also found that, as the number m of bits of X and Y increases, the difference in process time is remarkably increased. If a carry lookahead adder is used as the adder 411 shown in FIG. 4, the process time can be further shortened.
FIG. 5 shows a process of an alpha blending calculation in another example according to the invention. In this process, a complete alpha blending calculation is realized. In this example, it is assumed that each of X, Y, and .alpha. is 8-bit binary data. In FIG. 5, the meanings of X.sub.0 -X.sub.7, Y.sub.0 -Y.sub.7, .alpha..sub.0 -.alpha..sub.7, and P.sub.0 -P.sub.15 are the same as those described above, so that the descriptions thereof are omitted.
The alpha blending calculation shown in FIG. 5 is different from the alpha blending calculation shown in FIG. 3 in that (Y.sub.7, Y.sub.6, Y.sub.5, Y.sub.4, Y.sub.3, Y.sub.2, Y.sub.1, Y.sub.0) are additionally processed in the adding process in step 1. As a result, in step 1, nine stages of data are classified into three quotient groups g.sub.0, g.sub.1, and g.sub.2 by regarding three stages of data as a unit. Thereafter, in the same way as in the alpha blending calculation shown in FIG. 3, (1) the classification into quotient groups and a remainder group, and (2) the addition for respective digits of the quotient groups are repeatedly executed in each step. Finally, in step 5, the sum S.sub.6 and the carry C.sub.6 are added to each other considering the carry. As a result, a 16-bit fixed-point number P is obtained. The fixed-point number P indicates the result of the complete alpha blending calculation.
FIG. 6 schematically shows the organization of an alpha blending calculator which executes the alpha blending calculation in accordance with the procedure shown in FIG. 5. The alpha blending calculator includes a selector 600 for selecting one of the data X or Y in accordance with each bit of the digital data .alpha.. The selector 600 outputs Sel.sub.0 -Sel.sub.7 indicative of the selection results.
The alpha blending calculator further includes first level to fourth level adding portions 601 to 604, and an adder 612. The first level to fourth level adding portions 601 to 604 execute processes corresponding to steps 1 to 4 in FIG. 5, respectively. The adder 612 adds the sum S.sub.6 and the carry C.sub.6 output from the fourth level adding portion 604 considering the carry, and outputs the addition result as the 16-bit fixed-point number P.
The first level adding portion 601 is implemented by an adding portion 605 having a half adder (HA) and seven full adders (FA's), an adding portion 606 having two half adders (HA's) and six full adders (FA's), and an adding portion 607 having two half adders (HA's) and six full adders (FA's).
The second level adding portion 602 is implemented by an adding portion 608 having a half adder (HA) and seven full adders (FA's) and an adding portion 609 having five half adders (HA's) and five full adders (FA's).
The third level adding portion 603 is implemented by an adding portion 610 having four half adders (HA's) and six full adders (FA's).
The fourth level adding portion 604 is implemented by an adding portion 611 having five half adders (HA's) and seven full adders (FA's).
In the alpha blending calculator shown in FIG. 6, the number of levels of addition before the adder 612 is 4. In a conventional alpha blending calculator of the organization of array multiplier, the number of levels of the same addition is 7 (=8-1). Thus, according to the alpha blending calculator in this example, the number of levels of addition can be greatly reduced, so that the process time can be shortened. The conventional alpha blending calculator of the organization of array multiplier includes 7 half adders and 49 full adders. On the other hand, the alpha blending calculator shown in FIG. 6 includes 20 half adders and 44 full adders. Thus, the circuit scale of the alpha blending calculator in this example is slightly larger than that of the conventional one.
For simplifying the description, the invention is described by way of exemplary cases where each of the digital data X, Y, and .alpha. is composed of 4 bits and where each of the digital data X, Y, and .alpha. is composed of 8 bits. It is appreciated that the number of bits is not limited to the above-mentioned specific values. As described above, as the number of bits of the digital data increases, the reduction effect in process time is remarkably increased. In the field of digital images, a technique in which full-color images are displayed by using 24-bit data is developed. The present invention is suitable for executing the alpha blending calculation for digital data composed of such a large number of bits at a high speed.
In addition, the present invention can be applied to the case where the bit width of the digital data X and Y is not equal to the bit width of the digital data .alpha..
Various other modifications will be apparent to and can be readily made by those skilled in the art without departing from the scope and spirit of this invention. Accordingly, it is not intended that the scope of the claims appended hereto be limited to the description as set forth herein, but rather that the claims be broadly construed.

Number	Name	Date
4594678	Uhlenhoff	Jun 1986
4852037	Aoki	Jul 1989
5113363	Orsino et al.	May 1992
5285403	Quisquater et al.	Feb 1994

Number	Date	Country
6-83852	Mar 1994	JPX
9118355	Nov 1991	WOX

Alpha blending calculator

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 (4)

Foreign Referenced Citations (2)

Non-Patent Literature Citations (3)

Entry
M. Mehta et al., "High-Speed Multiplier Design Using Multi-Input Counter and Compressor Circuits", Proceedings of the 10th IEEE Symposium on Computer Arithmetic, pp. 43-50, (Jun. 26-28, 1991).
K. Lowe et al., "High-Speed GaAsa 4.times.4-Bit Parallel Multiplier Using Super Capacitor FET Logic", Electronics Letters, pp. 425-426, vol. 23, No. 8 (Apr. 1987).
R. White, et al., "Digital Filter Realizations Using a Special-Purpose Stored-Program Computer", IEEE Transactions on Audio and Electroacoustics, pp. 289-294 vol. AU-20, No. 4, (Oct. 1972).