JP4034385B2 - Multi-color image encoding apparatus and method, and multi-color image decoding apparatus and method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to a multi-color image encoding apparatus and method and a multi-color image decoding apparatus and method. More specifically, the present invention relates to an improvement in how to divide the target image, particularly when encoding and decoding a multi-color image.
[0002]
[Prior art]
Conventionally, an image called a multi-color image has been used in personal computers and game machines. This multi-color image is also called a representative color image or a limited color image. As shown in FIG. 22, a specific color, that is, a specific R (red), G (green), B (blue) ) Is an image in which an index is assigned to a color having a value of), and data is expressed by limited representative colors such as 16 colors or 256 colors using the data of the index.
[0003]
For such multi-color image data, if each color of R, G, and B is represented by 8 bits (256 types), a total of 24 bits are required, but the index itself is displayed by, for example, 8 bits. Therefore, the compression rate is considerable. However, although it is compressed, it still has a large amount of information, so if it is processed as it is without any contrivance, the memory capacity becomes large and the communication speed becomes slow, which is not practical. Therefore, the compression technique is very important for multicolor images as well as other image data. In particular, since the number of colors of a multi-color image is limited, lossless encoding and decoding, that is, a reversible compression technique is required.
[0004]
On the other hand, in recent years, a technique using an entropy encoder and a decoder has attracted attention as one of data compression techniques. As one of the entropy encoding and decoding techniques, for example, there is a technique using arithmetic encoding and decoding techniques. An outline of this technique is described in, for example, Japanese Patent Laid-Open Nos. 62-185413, 63-74324, 63-76525, and the like.
[0005]
FIG. 16 shows a conventional multi-color image encoding system 50 and decoding system 60 using such a technique. The encoding system 50 includes a line buffer 51 and an entropy encoder 52. The input index data, that is, the color pixel data 100 </ b> A is input to the line buffer 51 and the entropy encoder 52. As shown in FIG. 17, the color pixel data 100A is raster-scanned and sequentially input as pixel data in the horizontal scanning order. As a method of creating the index data, that is, the color pixel data 100A, a method of assigning an index in the order of input colors is generally used, as shown in FIG. For example, there is a phenomenon in which the colors are substantially different between “1” and “2”, or the colors are similar even when the index numbers are far, for example, “100” and “200”. In order to avoid such a phenomenon, as shown in Japanese Patent Application Laid-Open No. 5-328142, there are also those in which serial numbers are assigned to those having similar colors.
[0006]
The line buffer 51 in the encoding system 50 creates reference pixel data A, B, C, and D for the encoding target pixel X from the already input color pixel data 100A as reference pixel generation means. That is, the line buffer 51 stores a history of n lines (there are many 1 to 5 lines) when scanning an image. Each time the color pixel data 100A of the encoding target pixel X is input, a series of pixel data composed of the immediately preceding pixel A and the surrounding pixels B, C, and D is used as reference pixel data 110 as an entropy encoder. Output to 52.
[0007]
The entropy encoder 52 is formed using a technique such as arithmetic coding or Huffman coding, for example. Then, using the reference pixel data 110 as a status signal, the target color pixel data 100A is converted into encoded data 200 and output.
[0008]
On the other hand, the decoding system 60 includes a line buffer 61 and an entropy decoder 62. Here, the line buffer 61 and the entropy decoder 62 are formed so as to decode and output the input encoded data 200 in a procedure completely opposite to the line buffer 51 and the entropy encoder 52 of the encoding system 50. Yes.
[0009]
In this way, the encoding system 50 and the decoding system 60 encode the color pixel data 100A into the encoded data 200 using algorithms that are completely opposite to each other, and further convert the encoded data 200 into the color pixel data. 100B can be decoded and output. Therefore, this system can be widely used for various applications.
[0010]
By the way, in such a system, when the value of the color pixel data 100A, that is, the index number is biased near a certain number, the compression rate of the data is improved. In this system, the reference pixel data 110 is used as a status signal for the entropy encoder 52 and the entropy decoder 62. Therefore, if the number of states, that is, the number of reference pixels is increased, the data compression rate is similarly improved. That is, when the entropy encoder 52 and the decoder 62 are configured using a technique such as arithmetic coding or Huffman coding, if there is a large bias in the probability of occurrence of 0 or 1 symbols, the data compression rate is reduced. Can be increased. This is because in the entropy encoding technique, short encoded data is assigned to input data with a high occurrence probability, and relatively long encoded data is assigned to input data with a low occurrence rate.
[0011]
In order to obtain a large bias in the occurrence probability of symbols, that is, index numbers, conventionally, input data is classified into several states and encoded. This is because a good compression rate cannot be obtained without classification. For example, in the conventional method shown in FIG. 16, the line buffer 51, 61 is used to create reference pixel data, which is input to the entropy encoder 52 and the entropy decoder 62 as a classification state signal. The entropy encoder 52 and the entropy decoder 62 classify input data by using the state signal, and perform encoding and decoding. That is, the occurrence probability of each state of the reference pixel data is calculated, and short encoded data is assigned to a combination having a high occurrence probability. As a result, the data compression rate is increased.
[0012]
However, the entropy encoder 52 and the entropy decoder 62 described above require a number of encoding parameter tables corresponding to the number of states of the reference pixel data. For this reason, the larger the number of reference pixels in order to increase the compression ratio, the larger the parameter table for encoding and decoding. For this reason, there is a problem that the entropy encoder 52 and the entropy decoder 62 become large and expensive.
[0013]
For example, it is assumed that the color pixel data, that is, the index number is composed of 4-bit data (16 types) and the number of pixels of the reference pixel data 110 is four. In this case, the number of states of the encoding and decoding parameter table is 4 pixels × 4 bits = the number of states for 16 bits, that is, 2 ¹⁶ Take the number of states. For this reason, 2 ¹⁶ = 65536 parameter tables must be prepared. For this reason, each time the reference pixel is increased by one, the encoding and decoding parameter table is correspondingly increased, and the hardware configuring the entropy encoder 52 and the entropy decoder 62 is increased in size. Understood. In addition, the target pixel is also composed of 4 bits, that is, 4 planes, and a signal of 1 bit is assigned to each plane. As a result, 16 bits (colors) are obtained with 4 bits, so the parameter table is 65536 × 16. (See FIG. 18).
[0014]
In order to solve such a problem, the color symbol conversion method of calculating the color frequency deviation of the target pixel, that is, the index number corresponding to the color, and rearranging the index numbers corresponding to the appearance frequency order ( JP-A-6-276041). That is, by this, short encoded data is assigned to those with a high appearance frequency, and the compression rate is further increased. This publication also discloses a technique for reducing the parameter table according to the number of states degenerated in the entropy encoder 52 and the entropy decoder 62.
[0015]
The feature of the system for reducing the number of states disclosed in Japanese Patent Laid-Open No. 6-276041 is that, as shown in FIG. 19, the entropy encoder 52 and the entropy are the same as in the conventional encoding system 50 and decoding system 60. The reference pixel data 110 is input to the decoder 62 as a state signal. At the time of input, the state signal 140 is converted from the state signal degenerater 53, which degenerates the reference pixel data 110 output from the line buffers 51 and 61. It is in the point generated by 63.
[0016]
The state degeneration units 53 and 63 are configured to degenerate the input reference pixel data 110 into a state signal 140 having a smaller number of bits and output it to the corresponding entropy encoder 52 and entropy decoder 62. Yes. Each of the predictors 54 and 64 has, in its memory, a color order table (details will be described later) for converting the color pixel data into the color order based on the appearance frequency of the color symbols and vice versa. It is what.
[0017]
Note that degeneration is an operation of classifying the original state into the number of states after degeneration. This classification is performed by selecting the combination so that entropy (average information amount for displaying one symbol) after classification is minimized. Then, an identification bit is added to the number of states after degeneration, that is, the number of states after classification. This is the status signal 140.
[0018]
By the way, as a reduction table used for the state reduction units 53 and 63, a reduction table for specifying the relationship between the color symbol combination pattern of the reference pixel data 110 and the reduction data is set, and this reduction table is used for input. There is a method in which a combination pattern of color symbols in the reference pixel data 110 is converted into degenerate data and output.
[0019]
FIG. 20 shows an example of the degeneration operation performed using such a method. Here, in order to simplify the description, as shown in FIG. 20A, for the encoding target pixel X, a Markov model formed of three pixels A, B, and C is used as a reference pixel pattern. Will be described as an example.
[0020]
When the reference pixel is composed of three pixels as shown in FIG. 20A, there are five color symbol combination patterns as shown in FIG. 20B. That is, the pattern is classified into five patterns: a pattern in which the color symbols of all three pixels match, a pattern corresponding to the case in which only two color symbols match, and a pattern in which the color symbols of all pixels are different. The
[0021]
Therefore, by using the table shown in FIG. 20B as the degeneration table of the state degeneration units 53 and 63, a combination of three pixels can be originally obtained. ¹² The pattern state can be reduced to five states S1 to S5 shown in FIG. By doing so, the reference pixel data 110 can be effectively degenerated, and the number of states of the entropy encoder 52 and the entropy decoder 62 can be significantly reduced.
[0022]
By the way, a general technique of such arithmetic coding and decoding has already been described in detail in p26 to 44 and p44 to p50 of 1 image coding standard JBIG (International Standard ISO / IEC11544). Then, it will be briefly described as a prerequisite technique for developing the present invention described later.
[0023]
An example of the arithmetic code type entropy encoder 52 used in FIG. 16 is shown in FIG. Note that the configuration of the arithmetic decoding type entropy decoder 62 is substantially the same as the configuration of the entropy encoder 52, and thus the description thereof is omitted here.
[0024]
The entropy encoder 52 includes an arithmetic operation unit 55 and occurrence probability generation means 56 that functions as a state memory. In the occurrence probability generation means 56, a state parameter table necessary for determining the symbol occurrence probability required for encoding is written. The state parameter is specified by an input state signal. Then, the occurrence probability calculation parameter of the occurrence probability generation means 56 is output to the arithmetic operation unit 55 with respect to the state parameter table specified by the state signal. The arithmetic operation unit 55 performs entropy encoding based on the occurrence probability input in this way, and converts the input color rank data 120 into encoded data 200 and outputs it. Then, the occurrence probability for the state signal is recalculated based on the value of the encoded color rank data 120, and is input to the occurrence probability generation means 56 as an operation parameter update value. By storing this update result in the table as the occurrence probability of the next data, the compression efficiency of the entropy encoder 52 is improved.
[0025]
In order to generate the color order data 120, the color order table is arranged in the predictors 54 and 64 as described above. As an example of this color order table, one shown in FIG. 23 is known (see Japanese Patent Laid-Open No. 6-276041). In this example, when determining the color order table for the encoding target pixel X, the two-dimensional peripheral pixel data R0, R1, R2, and R3 are used as upper color order data, and the encoding target pixel X and A one-dimensional table of the same line is used as lower color order data. At this time, after the color symbols of the peripheral pixel data R0, R1, R2, and R3 are removed from the one-dimensional table, the upper and lower color order data are docked to form a color order table for the encoding target pixel X.
[0026]
A specific example of how the color order table is created will be described with reference to FIG. Consider a case where 16 color symbols are encoded. If the color order is fixed at the positions R0, R1, R2,... R8... Of each pixel as shown in FIG. 23 (A), the respective color symbols are C4, C3 as shown in FIG. , C6, C5, C2, C2,..., The color rank table that becomes the latest appearance table is as shown in FIG. That is, the highest level is C4 of R0, the second is C3 of R1, the third is C6 of R2, the fourth is C5 of R3, the fifth is C2 of R4, the sixth is C2 in R5, but C2 is Since it has already occurred and C4 of R6 has already occurred, the sixth is C0 of R7. In this way, the color order data excluding the color symbols already appearing in the upper rank, that is, the color symbols appearing in R0 to R3, are added to the data of R0 to R3, and the first to 16th color symbols of 16 colors are determined. . At this time, the upper four surrounding pixels can be made variable by learning.
[0027]
[Problems to be solved by the invention]
The color order table for performing the two-dimensional color order conversion shown in FIG. 23 deletes from the one-dimensional table the overlapping color symbols of the two-dimensional peripheral pixels and the one-dimensional table when the table is created. Work is needed. This process of deleting duplicate color symbols from the one-dimensional table increases the amount of processing as the number of bits of the color symbol index code increases. That is, as shown in FIG. 23, if the number of all color symbols is 16 colors, that is, if 4 bits are sufficient as an index code, duplicate color symbols are searched and the corresponding color symbols are deleted from the one-dimensional table. However, when the index code is 8 bits, there are 256 colors, and the operation of searching for and deleting duplicate color symbols becomes enormous.
[0028]
For example, when the peripheral pixel data R0, R1, R2, and R3 are all different colors, the four colors are searched from 256 colors in the one-dimensional table, and when the same color is found, the color symbol is deleted, Subsequent ranks are narrowed down and the 256th rank is assigned to the last color symbol. For this reason, a maximum of 256 × 4 comparisons must be made. This amount of work is extremely enormous. Further, in the one-dimensional table constituting the color order table shown in FIG. 23 and the color order table comprising only the one-dimensional table as shown in FIG. 45 of Japanese Patent Laid-Open No. 6-276041, a FIFO comprising a plurality of registers is used. In other words, the latest appearance table creation process of the so-called move to front process is performed. In the sorting process when creating the latest appearance table, the calculation amount increases as the number of bits of the index code increases, and the processing amount increases. For this reason, when an index code having a large number of bits is handled, the processing speed at the time of encoding or decoding a multi-color image is reduced.
[0029]
The line buffer 51 of the encoding system 50 and the line buffer 61 of the decoding system 60 as shown in FIGS. 16 and 19 can store a large number of pixels in order to generate reference pixels and peripheral pixels. ing. For this reason, the line buffers 51 and 61 are enlarged, and as a result, the apparatus is enlarged and expensive. Also, in the conventional apparatus shown in FIGS. 16 and 19, since the image is compressed or decoded with a fixed length in the horizontal direction, an arbitrary image size cannot be efficiently compressed or decoded. It has become.
[0030]
The present invention has been made in response to the above problems, and a multicolor image encoding apparatus and method capable of using peripheral pixel generation means such as a low-capacity line buffer for encoding and decoding. Another object of the present invention is to provide a multicolor image decoding apparatus and method.
[0031]
[Means for Solving the Problems]
In order to achieve such an object, according to the first aspect of the present invention, in a multi-color image encoding device that encodes input target color pixel data into encoded data and outputs the encoded data, the target image to be encoded is detected. A strip generation means for outputting pixel data corresponding to the strip, and a two-dimensional peripheral pixel corresponding to the pixel data based on the pixel data corresponding to the strip. Peripheral pixel data and encoding target pixel 4 surrounding reference pixels excluding the previous pixel Line buffer means for generating reference pixel data for and outputting the peripheral pixel data and reference pixel data, creating a color order table based on the peripheral pixel data, and based on the reference pixel data A Markov model context is created, and the encoding target pixels located in the strips are sequentially encoded based on the color order table and the context.
[0032]
Thus, since the target image is divided into strips having a limited number of pixels, the circuit scale and the amount of memory can be reduced when generating peripheral pixels and reference pixels using a line buffer that embodies the line buffer means. . As a result, the apparatus can be reduced in size and cost.
In addition, when encoding multicolor images, create a context for the Markov model. 4 surrounding pixels excluding the previous pixel Therefore, the encoding process can be speeded up.
[0033]
Furthermore, in the second aspect of the present invention, in the multi-color image encoding method that encodes the input target color pixel data into encoded data and outputs the encoded data, the target image to be encoded has a limited number of pixels. A strip generation step for outputting pixel data corresponding to the strip by dividing into strips having a horizontal width, and peripheral pixel data and encoding for a two-dimensional peripheral pixel corresponding to the pixel data based on the pixel data corresponding to the strip Of the target pixel 4 surrounding pixels excluding the previous pixel A line buffer step for generating reference pixel data for the reference pixel and outputting the peripheral pixel data and the reference pixel data, and creating a color order table based on the peripheral pixel data, and the reference pixel A Markov model context is created based on the data, and the encoding target pixels located in each strip are sequentially encoded based on the color order table and the context.
[0034]
In this way, the target image is divided into strips having a limited width of the number of pixels, so the circuit scale and memory amount of the line buffer used in the peripheral pixel generation process and the reference pixel generation process used in the line buffer process, etc. Can be reduced. As a result, it is possible to reduce the size and cost of an apparatus that embodies this method.
In addition, when encoding multicolor images, create a context for the Markov model. 4 surrounding pixels excluding the previous pixel Therefore, the encoding process can be speeded up.
[0035]
Furthermore, in the invention described in claim 3, in a multi-color image decoding apparatus that decodes input target encoded data into decoded data and outputs the decoded target data, the target image to be decoded has a limited number of pixels. A strip compositing means for dividing and decoding into strips having a width, and generating two-dimensional peripheral pixel data and reference pixel data corresponding to the pixel data based on the pixel data corresponding to the strip, and the peripheral pixel data and the peripheral pixel data Line buffer means for outputting reference pixel data, and creating a color order table based on the peripheral pixel data; 4 surrounding pixels excluding the previous pixel Based on the decoded data, a Markov model context is created, and based on the color order table and the context, the target images to be decoded are sequentially decoded in a strip shape.
[0036]
In this way, the target image is divided into strips having a limited width of the number of pixels, so that the circuit scale and memory amount when generating peripheral pixels and reference pixels with a line buffer that embodies the line buffer means are reduced. it can. As a result, the apparatus can be reduced in size and cost.
In addition, when decoding multicolor images, create a context for the Markov model. 4 surrounding pixels excluding the previous pixel Therefore, it is possible to increase the speed of the decoding process.
[0037]
In addition, in the invention according to claim 4, in the decoding method of the multi-color image in which the input target encoded data is decoded into the decoded data and output, the target image to be decoded is limited in the number of pixels. A two-dimensional peripheral pixel data and reference pixel data corresponding to the pixel data based on the strip composition step for decoding the strip into a strip having a horizontal width and decoding, and the pixel data corresponding to the strip, and the peripheral pixel data and A line buffer step for outputting the reference pixel data, and creating a color order table based on the peripheral pixel data; 4 surrounding pixels excluding the previous pixel Based on the decoded data, a Markov model context is created, and based on the color order table and the context, the target images to be decoded are sequentially decoded in a strip shape.
[0038]
In this way, the target image is divided into strips having a width of a limited number of pixels, so the circuit scale and memory amount of the line buffer used for peripheral pixel generation and reference pixel generation used in the line buffer process are reduced. it can. As a result, it is possible to reduce the size and cost of an apparatus that embodies this method.
In addition, when decoding multicolor images, create a context for the Markov model. 4 surrounding reference pixels excluding the previous pixel Therefore, the decoding process can be speeded up.
[0065]
DETAILED DESCRIPTION OF THE INVENTION
An embodiment of the present invention will be described with reference to FIGS. In addition, the same code | symbol is attached | subjected and demonstrated to each data corresponding to the data in a prior art.
[0066]
FIG. 1 shows a preferred embodiment of a multi-color image encoding system 1 according to the present invention. FIG. 2 shows a preferred embodiment of a multi-color image decoding system 3 corresponding to the encoding system 1 of FIG.
[0067]
The encoding system 1 includes a strip generation unit 5, a line buffer 10 serving as a peripheral pixel generation unit and a reference pixel generation unit, a Markov model generation unit 11 serving as a degeneration unit, and an entropy encoder 12 serving as an entropy coding unit. And a predictor 13 serving as a predicting unit, and is configured to convert an input data stream of color pixel data 100A into a data stream of encoded data 200 and output the data stream.
[0068]
The predictor 13 includes a color rank generation unit 14 and a determination unit 15 that also serves as an escape code generation unit. The color order generation unit 14 also receives the outputs from the two-dimensional color order generation unit 16 that determines the upper color order, the one-dimensional color order generation unit 17 that determines the lower color order, and the generation units 16 and 17. The combined color order table 23 and the generated prediction table synthesis unit 18 are configured. Further, the two-dimensional color order generation unit 16 includes a priority order switching unit 19 and a two-dimensional color order table 20, and the one-dimensional color order generation unit 17 includes a table update unit 21 and a one-dimensional color order table 22. It consists of and.
[0069]
In this embodiment, the color pixel data 100A to be encoded is a multi-color pixel, is composed of 8-bit index code data per pixel, and can display color symbols for 256 colors.
[0070]
The input color pixel data 100 </ b> A is supplied to the line buffer 10 and the table updating unit 21 in the predictor 13.
[0071]
As shown in FIG. 3, the line buffer 10 has a function of inputting and storing pixel data of a portion obtained by dividing a target image into strips having a limited number of pixels. Specifically, an entire image as shown in FIG. 3 is taken into a memory (not shown), and then the memory is accessed to take out data. At that time, the data is extracted in the shape of strips L1 to L5 shown in FIG. The line buffer 10 stores at least the 17 most recently input pixels in each of the strips L1 to L5 divided into 16 horizontal pixels. Then, the value is input as a peripheral pixel to the two-dimensional color rank generation unit 16 to create a two-dimensional color rank table 20, and is also input as a reference pixel to the Markov model generation unit 11 to generate a state signal Cx.
[0072]
As shown in FIG. 4, the Markov model generation unit 11 has four pixels (A, B, C, D) around the pixel X to be encoded, excluding the immediately preceding pixel (indicated by “−”) as a reference pixel. And the state signal Cx is generated. Here, without adopting the previous pixel, 2 The reason why A, which is the pixel before the pixel, is used as the reference pixel is as follows. When the immediately preceding pixel is adopted as a reference pixel as in the conventional case, the state signal Cx input to the entropy encoder 12 or the entropy decoder 32, that is, the context of the Markov model, is not determined until the color symbol of the pixel is determined, and encoding is performed. And the decryption processing operation cannot be accelerated. In this embodiment, A, which is a pixel two pixels before, is used as a reference pixel in order to increase the speed.
[0073]
The entropy encoder 12 is an arithmetic entropy encoder shown in FIG. That is, it has a conversion table (not shown) for each state parameter therein.
[0074]
The predictor 13 functions as a prediction unit. Then, the color pixel data 100A to be encoded is input to the color order generation unit 14 and the determination unit 15 in the predictor 13, respectively. In addition, peripheral color data is input from the line buffer 10 to the color order generation unit 14.
[0075]
The peripheral pixel data input from the line buffer 10 is input to the priority switching unit 19. The peripheral pixel data is pixels P0, P1, P2, and P3 around the encoding target pixel X as shown in FIG. Here, the reference pixels A, B, C, and D of the Markov model shown above are B = P1, C = P2, and D = P3, and only the pixel A is different.
[0076]
The priority switching unit 19 is configured to change the priority for generating the two-dimensional color order table 20 based on a predetermined switching command. However, in this embodiment, the order and position of P0, P1, P2, and P3 are fixed as shown in FIGS. As a method of switching without fixing, for example, pre-scanning is performed on an image to be encoded, and a priority pattern of peripheral pixels that can obtain a good compression rate is obtained in advance, or encoding or decoding is performed. It is possible to employ a method of calculating the number of times of matching color symbols during the operation and outputting a switching command to bring the pixel at the position with the highest number of matches to the upper level.
[0077]
Based on the command from the priority switching unit 19, a two-dimensional color order table 20 (see FIG. 6) is created, which is a higher rank of the color order table in the prediction table synthesis unit 18. The color order table in the prediction table synthesizing unit 18 is the so-called latest appearance table as shown in FIG. 8, and the upper four levels in this embodiment, the highest four from the highest 0 to the third in this embodiment, are two-dimensional colors. It is obtained from the ranking table 20. The two-dimensional color order table 20 is stored in a memory (not shown).
[0078]
When the color pixel data 100A is input to the table update unit 21 in the one-dimensional color order generation unit 17, the table update unit 21 searches the one-dimensional color order table 22 stored in the memory to input color pixels. The latest appearance table creation operation for raising the color rank of the color symbol corresponding to the data to the 0th place is performed. For example, when the color symbol C4 at the position of T0 in FIG. 5 is input, the table updating unit 21 searches the one-dimensional color order table 22, extracts the corresponding color symbol C4, and sets the color order of the C4 to the first. It moves up to 0th place, C2 which was 0th place before is moved to 1st place, C0 which was 1st place is moved down to 2nd place in order. Then, the updated one-dimensional color order table 22 is stored in a memory (not shown).
[0079]
The one-dimensional color order generation unit 17 may be composed of a FIFO unit composed of 256 registers, but in this embodiment, in order to improve the processing speed, the upper rank from the 0th place to the 31st place Only 32 in total are composed of FIFO units composed of registers, and a complete latest appearance table creation operation is performed. On the other hand, the 32nd to 255th positions are composed of a 224 byte dual port RAM, and when the corresponding color symbol is the nth place below the 32nd place, the corresponding color symbol is moved up to the 0th place, A process of replacing the original 31st place (= 32nd place) with the nth place is performed. The reason why the 32nd and lower ranks are composed of dual-port RAMs is to increase the processing speed by parallel processing, and in particular when considering cost reduction and downsizing, they should be composed of single-port RAMs. May be.
[0080]
The prediction table synthesizing unit 18 receives the data of the two-dimensional color order table 20 and the data of the one-dimensional color order table 22, and creates a color order table 23 as a kind of latest appearance table shown in FIG. The As shown in FIG. 8, the color order table 23 is composed of peripheral pixels corresponding to the two-dimensional color order table 20 in the maximum 0th to 3rd positions. For example, when the colors of P0, P1, P2, and P3 of the peripheral pixels are all different, those pixels occupy from the 0th place to the 3rd place. On the other hand, when all of P0 to P3 have the same color, only the 0th place is occupied, and the first place in the one-dimensional color order table 22 is also the first place in the color order table 23.
[0081]
In the color order table 23, the 0th place is the 0th place in the two-dimensional color order table 20 and the 0th place in the one-dimensional color order table 22. In addition, an escape code is attached as a code to the 255th place, that is, the order of 256th and later from the top. Specifically, at the 255th place, an escape code represented by “11111111” and a code indicating “0” represented by “00000000” are output. Similarly, an escape code of “11111111” and a code indicating “1” of “00000001” are output to the 256th place.
[0082]
As described above, the synthesized color order table 23 is configured not to delete the overlapping portion between the values of the two-dimensional color order table 20 and the one-dimensional color order table 22 which are the values of surrounding pixels. A search (comparison) operation is not required, and the speed is increased. The deterioration of the compression rate is 1% or less compared to the conventional one, while the processing speed is more than twice as fast. In addition, it is possible to respond by increasing the number of bits of the code of the color order without using such an escape code. However, in this case, the efficiency of encoding is greatly deteriorated due to the increase in the number of bits. End up. On the other hand, the use of the escape code is extremely advantageous because the compression rate hardly deteriorates.
[0083]
The combined data of the color order table 23 is output to the determination unit 15. The discriminating unit 15 collates the data of the color order table 23 combined with the color symbol of the color pixel data 100A to be encoded, and obtains the color order data 120 corresponding to the color pixel data 100A to be encoded. Output.
[0084]
The entropy encoder 12 receives the previous status signal Cx and the color rank data 120 and outputs the encoded data 200.
[0085]
Next, the operation of the encoding system 1 having the above configuration will be described based on the flowchart shown in FIG. In the description, first, an operation in a normal case will be described, and then an exception process at the end of each of the strips L1 to L6 will be described.
[0086]
In encoding, first, a strip generation process for extracting data from the memory into a strip shape is performed. Thereafter, the encoding of each pixel is specifically started. In the normal encoding process, the color pixel data 100A two before the encoding target pixel has already been input (step S1). A Markov model context is created by the Markov model creation unit 11 from this data and data of three reference pixels on one line (step S2).
[0087]
In parallel with this context creation processing, the following processing is performed for color order conversion. That is, the data one pixel before the pixel to be encoded has also been input (step S3), and the input of all the peripheral pixels is completed by the input of the pixel (step S4), and the two-dimensional color order table 20 is created ( Step S5). On the other hand, the input of data one pixel before the encoding target pixel is the same as the previous pixel input (step S6) to the one-dimensional color order generation unit 17, and the one-dimensional color order table 22 is created by the input. (Step S7).
[0088]
Next, the respective data of the two-dimensional color order table 20 and the one-dimensional color order table 22 are input to the prediction table composition unit 18, and the synthesized color order table 23 is created (step S8). At this time, numbers (0 to 3) different from the escape code (= “11111111”) are all assigned to the color symbols of the 255th (256th) and lower ranks.
[0089]
Next, the color pixel data 100A to be encoded is input (step S9) and collated with the data of the synthesized color order table 23 (step S10). When the color pixel data 100A is different from that in the two-dimensional color order table 20, the part corresponding to the one-dimensional color order table 22 is searched to find the corresponding color symbol. For example, if the color symbol is in the 10th position (11th) in the one-dimensional color order table 22, and there are four orders in the two-dimensional color order table 20, the 3 is added, The color rank data 120 which is the 13th place is output (step S11). In this way, the substantial search is performed on the two-dimensional color order table 20 and the one-dimensional color order table 22, and when found in the two-dimensional color order table 20, the searched order is output as it is. If it is in the one-dimensional color order table 22, a predetermined value is added and output.
[0090]
On the other hand, the corresponding color symbol is brought to the 0th place of the one-dimensional color order table 22, and the color symbol is brought to the 0th place of the two-dimensional color order table 20 and the combined color order table 23. Such updating of the tables 20, 22, 23 is performed in step S12. At this time, in the lower part of the one-dimensional color order table 22 and the synthesized color order table 23, when the value is from the 0th place to the 31st place, a complete latest appearance table creation operation is performed, Things go up to the 0th place, others go down one by one. However, when the number 32 or lower is the target, as shown in FIG. 9, the replacement process is performed to bring the original number 31 to the corresponding rank.
[0091]
The entropy encoder 12 performs encoding processing from the output color rank data 120 and the Markov model context, that is, the state signal Cx (step S13), and outputs the encoded data 200 (step S14). Then, it is determined whether or not the encoding is finished (step S15). If it is finished, the coding is finished, and if it is continued, the process returns to step S9.
[0092]
The above is the normal encoding step. Next, exception processing at the end of each strip L1 to L5 will be described.
[0093]
First, the case where the first portion of the first strip L1, that is, the leftmost pixel at the top of FIG. 11A is encoded will be described. First, target encoding target color pixel data 100A is input. That is, step S9 is executed first. At this time, since the peripheral pixels are not input, the two-dimensional color order table 20 is initialized to the zeroth one of the one-dimensional color order table 22, while the one-dimensional color order table 22 is also 256 colors. The color symbols are arranged from the 0th place to the 255th place and are initialized. For this reason, as the synthesized color order table 23, a table of 256 color symbols arranged from the 0th place to the 255th place exists in the prediction table composition unit 18. The line buffer 10 is also initially initialized to “0”. Note that a color symbol having a value other than “0” may be employed as the initial value of the line buffer 10.
[0094]
Thereafter, steps S10, S11, S12, S13, S14, and S15 shown in FIG. 10 are performed, and the process returns to step S9. At this time, in the update of the two-dimensional color order table 20 in the update of each table (step S12), a process of inserting a color symbol corresponding to the color pixel input at the position P0 of the two-dimensional color order table 20 is performed. Since P1, P2, and P3 have not yet occurred, the initial value “0” of the line buffer 10 becomes a color symbol. On the other hand, in the one-dimensional color order table 22, a leading movement process for moving up the color symbol corresponding to the color pixel input at the 0th position, that is, the latest appearance table creation operation is performed.
[0095]
In the uppermost line of the strip L1 shown in FIG. 11B, only P0 is always input as the peripheral pixel. On the other hand, the one-dimensional color rank table 22 always performs the latest appearance table creation operation, and performs the process of raising the most recently input color symbol to the top. Note that the original Markov model context is not created when the first pixel is input, and the status signal Cx outputs Cx = 0 as one color as a temporary value. The same applies to the input of the second pixel, and the value A shown in FIG. 4 is determined from the input time of the third pixel, but the original Markov model cannot be formed. However, since the color symbol of the pixel on one line, that is, the index is initialized to “0”, if the color symbol of P0 is “0”, the status signal is output as Cx = 0 indicating one color If the color symbol of P0 is not “0”, Cx = 1 indicating two colors is output as the status signal.
[0096]
Thereafter, when the color pixel data 100A at the left end (17th pixel) of the second line of the strip L1 is input, P0, P1, and P2 as peripheral pixels are as shown in FIG. That is, the pixel one pixel before the encoding target pixel X is P0, the pixel 15 pixels before is P2, and the pixel 16 pixels before, that is, the same position on one line is P1. Since P3 has not yet occurred, processing is performed assuming that the color symbol “0” is contained.
[0097]
As described above, in this embodiment, the peripheral pixels P0 to P3 and the reference pixels A to D are set to values at positions away from the encoding target pixel X within the line buffer 10 by a certain distance. That is, as shown in FIG. 12B, the peripheral pixel P0 is one previous, the reference pixel A is two previous, the peripheral pixel P2 (which is also the reference pixel C) is fifteen, the peripheral The pixel P1 (which is also the reference pixel B) is 16 previous, that is, the same position one line before the encoding target pixel, and the peripheral pixel P3 (which is also the reference pixel D) is 17 previous. The concept described above is the same when the encoding target pixel X, peripheral pixels, and the like are separated into strips L2 and strips L1, as shown in FIG. In this way, if 17 pixels are stored in the line buffer 10, they can always be used as reference pixels and peripheral pixels. Further, although a simple operation of setting “0” as the initial value of the line buffer 10 is performed, troublesome exception processing such as virtually inserting a value of “0” is not performed during the encoding. . That is, in this embodiment, it is possible to proceed from one strip to the next strip with the same processing operation as usual without specially performing troublesome exception processing.
[0098]
Further, when the strip is divided in units of 16 pixels in width, the width of the last strip L5 may be less than 16 pixels as shown in FIG. In this case, when the width of the strip L5 is 4 pixels or more, as shown in FIG. 13, if the number of pixels of the width is n, P0 is one pixel before the encoding target pixel and the reference pixel A is two pixels before The peripheral pixel P2 (reference pixel C) is (n−1) previous, the peripheral pixel P1 (reference pixel B) is n previous, and the peripheral pixel P3 (reference pixel D) is (n + 1) previous.
[0099]
When the width of the strip L5 is 3 pixels, as shown in FIG. 14A, the reference pixel A is the same as the reference pixel C and becomes a place of the peripheral pixel P2. Otherwise, the relationship is the same as when the width is 4 pixels. If the width of the strip L5 is 2 pixels, as shown in FIG. 14B, the previous pixel X is P0, the second is P1, and the third is P3. Here, in the relationship shown in FIG. 13, P2 is (2-1) previous, that is, one before, and has the same value as P0. However, the reference pixel C corresponding to P2 is not the previous one, but is at the position of P1. For this reason, the reference pixels B, C, and A are all at the position P1. Further, if the width of the strip L5 is one pixel, the relationship of FIG. 13 is applied mutatis mutandis, and as shown in FIG. 14C, the previous pixel X to be encoded indicates P0, P1, and P2. The last two indicate P3. Here, the reference pixels A, B, C, and D are all the same and are at the position P3.
[0100]
When the exception processing as described above is performed, for example, the edge portion is processed by virtually putting “0” in a portion where the positional relationship between the peripheral pixel and the reference pixel is insufficient, or for a strip having a width of less than 16 pixels, The processing is more efficient than the case of processing to have a horizontal width of 16 pixels, and an image having a horizontal width can be easily handled.
[0101]
Next, a multi-color image decoding system 3 corresponding to the encoding system 1 will be described.
[0102]
The decoding system 3 includes a strip synthesizing unit 7, a line buffer 30 serving as a peripheral pixel generating unit and a reference pixel generating unit, a Markov model generating unit 31 serving as a degeneration unit, and an entropy decoder 32 serving as an entropy decoding unit. And a predictor 33 serving as a prediction means, and is configured to convert the input encoded data 200 data stream into a color pixel data 100B data stream and output the data stream. At this time, the algorithm of the decoding system 3 is configured to be exactly the reverse of the algorithm of the encoding system 1. Therefore, the bit configuration and data stream of the color pixel data 100A and the bit configuration and data stream of the color pixel data 100B are exactly the same. The predictor 33 has the same configuration as the predictor 13 in the encoding system 1. That is, the predictor 33 includes a color rank generation unit 14 and a determination unit 15 that also serves as an escape encoding generation unit. The color order generation unit 14 also receives the outputs from the two-dimensional color order generation unit 16 that determines the upper color order, the one-dimensional color order generation unit 17 that determines the lower color order, and the generation units 16 and 17. And a prediction table synthesizing unit 18 for generating a synthesized color order table 23. Further, the two-dimensional color order generation unit 16 includes a priority order switching unit 19 and a two-dimensional color order table 20, and the one-dimensional color order generation unit 17 includes a table update unit 21 and a one-dimensional color order table 22. It consists of and.
[0103]
The line buffer 30 can store at least 17 pixels as in the line buffer 10 of the encoding system 1. Then, the values are input as peripheral pixels to the two-dimensional color rank generation unit 16 to create the two-dimensional color rank table 20, and are also input as reference pixels to the Markov model generation unit 31 to generate the state signal Cx.
[0104]
The entropy decoder 32 uses the state signal Cx to decode the input encoded data 200 in the reverse procedure of the entropy encoder 12, converts this into color rank data 120, and converts it into the color rank data 120. Output to. It should be noted that the entropy decoder 32 must be formed so as to perform the operation with an algorithm completely opposite to that of the entropy encoder 12. Therefore, the entropy decoder 32 needs to be formed as an arithmetic decoder having the same configuration when an arithmetic encoder is used for the entropy encoder 12, and the entropy encoder 12 has a Huffman encoder. When used, it is necessary to configure as a Huffman decoder having the same configuration. As a result, the encoded data 200 is accurately converted into color order data 120 paired therewith and output.
[0105]
The predictor 33 functions as a prediction unit. Then, a synthesized color order table 23 that is the same as that of the predictor 13 of the encoding system 1 is set therein, and the color order table 23 corresponds to the input color order data 120 and the data of surrounding pixels. The color symbol to be decoded is decoded and output as color pixel data 100B.
[0106]
Next, the operation of the decoding system 3 configured as described above will be described based on the flowchart shown in FIG.
[0107]
First, the encoded data 200 of the decoding target pixel is input to the entropy encoder 32 (step S31). On the other hand, a Markov model context is created based on the data two pixels before (step S32), and is input to the entropy decoder 32 as the state signal Cx. The entropy decoder 32 decodes the encoded data 200 using the state signal Cx (step S33).
[0108]
Through the decoding process by the entropy decoder 32, the color rank data 120 is output (step S34) and input to the determination unit 15 in the predictor 33.
[0109]
On the other hand, data that has already been decoded as the color pixel data 100B is input to the priority switching unit 19 of the predictor 33 as peripheral pixels via the line buffer 30 (step S35) and at the same time, a table update unit in the predictor 33 21 is input as the immediately preceding pixel to step 21 (step S36). When the peripheral pixels are input, the two-dimensional color order table 20 is created (step S37). When the immediately preceding pixel is input, the one-dimensional color order table 22 is created (step S38).
[0110]
Here, the two-dimensional color order table 20 is the same as that shown in FIG. 6, and the one-dimensional color order table 22 is the same as that shown in FIG. Then, as shown in FIG. 9, the table update unit 21 is a FIFO unit that performs complete head movement operation from the 0th place to the 31st place, and its lower part is composed of a RAM of 224 bytes. ing.
[0111]
The prediction table synthesis unit 18 creates a color order table 23 synthesized from both tables 20 and 22 (step S39). The data of the color order table 23 is input to the determination unit 15 and collated with the previous color order data 120 (step S40). Here, the color order data is, for example, the code “00111111” for the 63rd place, and the escape codes “11111111” and “00000001” for the 256th place. As a result of the collation, the color pixel data 100B (= index code) of the color symbol at the corresponding rank in the combined color rank table 23 is output (step S41).
[0112]
Next, for example, when the code of the second place comes out, the color pixel data 100B of the corresponding color symbol is output, and the color symbol in the synthesized color order table 23 is moved up to the 0th place. Update processing is performed to lower the 0th one to the first and the first one to the second (step S42). This update is also performed as necessary for the two-dimensional color order table 20 and the one-dimensional color order table 22.
[0113]
In the update operation, the order of P0, P1, P2, and P3 is determined in the two-dimensional color order table 20, but what is decoded is the same as any one of P0, P1, P2, and P3. For example, if it is the same as P2 in the second place, the color symbol is moved up to the 0th place. At this time, if the leading movement process is not performed at all for the one-dimensional color order table 22, the one-dimensional color order table 22 at the time of encoding will be different. Therefore, the same color symbol as the color symbol that was in the second position in the two-dimensional color order table 22 is searched in the one-dimensional color order table 22, and the color symbol is raised to the zeroth position. The top move process to rearrange things is performed.
[0114]
At this time, the search of the one-dimensional color order table 22 can be limited to the top 17 searches from the 0th place to the 16th place in the one-dimensional color order table 22. That is, P0 to P3 are one of the latest 17 pixels in the one-dimensional color order table 22. In this way, if the pixel to be decoded is any one of P0 to P3, only a maximum of 17 searches (comparisons) are required for updating the one-dimensional color order table 22. As described above, the reason why the number of times is limited to 17 times is that the image is divided into strips having a width of 16 pixels, so that the number of reference pixels stored in the line buffer 10 can be limited to 17.
[0115]
On the other hand, when the pixel to be decoded is not any of P0 to P3, when the color rank data 120, for example, the color rank data 120 indicating the 150th rank is output in step S34, a total of four colors are obtained in P0 to P3. In this case, since the color order data 120 corresponds to the 153rd place in the one-dimensional color order table 22, the 153rd place color symbol is directly extracted without searching. It will be moved up.
[0116]
In this way, in the case of decoding, a maximum of 16 searches (comparisons) are performed. The number of searches (comparisons) of 16 times is the same as that of the conventional technique disclosed in Japanese Patent Laid-Open No. 6-276041. Therefore, the number of searches (256 × 4) is extremely small, and extremely high-speed processing is possible. The search at the time of decoding may be equivalent to that at the time of encoding, and the entire one-dimensional color order table 22 may be searched. However, the above-described embodiment is preferable in terms of high-speed processing. .
[0117]
When the color pixel data 100B is output, it is determined in step S43 whether or not decoding is completed. Then, when it is determined that there is no input of new encoded data 200 and decoding is completed, the process ends. On the other hand, when continuing, it returns to step S31. In the decoding process, exception processing relating to the end of the strip or the strip having a small width is executed by the same algorithm as shown in FIGS. The decoded color pixel data 100B is decoded into a strip shape by a strip decoding process by the strip synthesis means 7, and the data is taken into a memory (not shown) and decoded as an entire image as shown in FIG. The
[0118]
In this embodiment, two codes (16 bits in total), which are an escape code and codes indicating a numerical value obtained by subtracting 256 from the overall rank (those counted from the first), are those having a rank of 255th or lower. Although the output is low, a code with a low rank of 255th or lower indicates a color code that hardly occurs, and the output of such two codes hardly occurs. Moreover, even if it occurs, since it is assigned to a rank that is easy to appear, the compression rate hardly deteriorates as described above. Compared to this, the effect that the overlapping color code can be deleted and the sorting process for changing the order can be eliminated is extremely large.
[0119]
Unlike the conventional Markov model of this embodiment, the pixel immediately before the encoding target pixel is not adopted as the reference pixel, but the deterioration of the compression rate is suppressed to several percent. On the other hand, the processing speed can be pipelined between the entropy encoder 12 and the entropy decoder 32 and the color rank conversion, and the processing speed has been increased by more than twice. In this embodiment, the immediately preceding pixel is used as a peripheral pixel. However, unlike this embodiment, in the case where the immediately preceding pixel is not adopted as a peripheral pixel for creating the color order table 23. However, the deterioration of the compression rate is suppressed to about 10%, while the processing speed becomes twice or more.
[0120]
Furthermore, the line buffers 10 and 30 of this embodiment can have a small capacity because the images are partitioned into strips. That is, as the reference pixel of the Markov model and the peripheral pixel for color order conversion, it is necessary to see a pixel on one line, but since the horizontal width is sectioned in a strip shape, the pixel on one line is seen. The number of pixels that should be stored in is reduced. Therefore, the line buffers 10 and 30 can be reduced in size and capacity, and the cost can be reduced.
[0121]
The above-described embodiment is an example of a preferred embodiment of the present invention, but the present invention is not limited to this, and various modifications can be made without departing from the gist of the present invention. For example, the color pixel data 100A can be configured to encode and decode n-bit (n is an integer of 2 or more) color pixel data 100A.
[0122]
Further, the method of dividing the target image into strips can be applied to a conventional apparatus or method for deleting duplicate color symbols shown in FIG. In that case, the processing can be speeded up. Further, a normal encoding system 50 or decoding system 60 as shown in FIG. 16 which does not have the combined color order table 23 shown in this embodiment and the conventional combined color order table shown in FIG. A method of dividing the target image into strips can also be adopted. In this case, the line buffers 51 and 61 have effects such as a reduction in capacity and cost.
[0123]
Further, the width of the strip is preferably 8 to 64 pixels in consideration of the capacity of the line buffers 10 and 30 and the like, but 16 to 32 pixels is most preferable in consideration of the search efficiency and the compression efficiency due to the exception processing at the end. Further, as a configuration ratio of the FIFO unit and the RAM in the one-dimensional color order generation unit 17, it is preferable that the FIFO unit is 1/4 to 1/32 of the whole and the rest is a RAM, and the FIFO unit is 1/8 to It is most preferable to use 1/16 as the remaining RAM because the compression rate does not drop so much and high-speed processing is possible.
[0124]
As encoding and decoding means, in addition to the arithmetic code type entropy encoder 12 and the decoder 32, a run-length encoding (decoding) technique and other encoding (decoding) techniques are adopted. be able to.
[0125]
Further, as the context of the Markov model, other state signals Cx are generated such as changing the state signal Cx depending on where there are different colors as shown in FIG. 20 in addition to the four states of one to four colors. You may make it let it.
[0126]
Also, in order to make the reference pixel of the Markov model different from the peripheral pixel for color order conversion, the pixel before one pixel is not adopted as the reference pixel as shown in this embodiment, but the pixel before two pixels is adopted. In addition, it may be adopted as a reference pixel before three or more pixels. That is, in addition to the previous pixel, the previous two pixels may not be used as the reference pixel.
[0127]
This Markov model context creation technology, that is, the technology that does not employ the pixel immediately before the pixel to be encoded as the reference pixel is not parallel to the technology that performs the color order conversion, and is various encoding and decoding technologies independently. Can be used.
[0128]
【The invention's effect】
As described above, in the multi-color image encoding apparatus and method according to the present invention, by dividing the target image into strips, it is possible to significantly reduce the amount of calculation, speeding up the encoding process, and arbitrary images. Multi-color images of size can be encoded efficiently. In addition, it is possible to reduce the capacity and cost of peripheral pixel generation means used for encoding, such as a line buffer.
[0129]
In the multicolor image decoding apparatus and method according to the present invention, the amount of calculation can be greatly reduced by dividing the target image into strips, the decoding process can be speeded up, and a multicolor of an arbitrary image size can be obtained. Images can be encoded efficiently. In addition, it is possible to reduce the capacity and the cost of the line buffer serving as the peripheral pixel generating means used for decoding.
[Brief description of the drawings]
FIG. 1 is a diagram illustrating a multicolor image encoding apparatus and encoding system according to the present invention.
FIG. 2 is a diagram illustrating a decoding system employing the multi-color image decoding apparatus and method according to the present invention.
FIG. 3 is a diagram for explaining a scanning method employed in the encoding system and the decoding system of FIGS. 1 and 2;
FIG. 4 is a diagram for explaining a Markov model employed in the encoding system and the decoding system in FIGS. 1 and 2 and its context; FIG. 4A is a diagram showing an arrangement of reference pixels employed as the Markov model; (B) is a figure which shows the context of a Markov model, ie, the kind of state signal.
FIG. 5 is a diagram for explaining pixels used to create a color order table employed in the encoding system and the decoding system of FIGS. 1 and 2; FIG. It is a figure which shows a pixel priority, (B) is a figure which shows the color symbol of each pixel of (A).
6 is a diagram showing a two-dimensional color order table employed in the encoding system and the decoding system of FIGS. 1 and 2. FIG.
7 is a diagram showing a one-dimensional color order table employed in the encoding system and the decoding system of FIGS. 1 and 2. FIG.
8 is a diagram showing a combined color order table employed in the encoding system and decoding system of FIGS. 1 and 2. FIG.
FIG. 9 is a diagram for explaining a configuration of a one-dimensional color order table employed in the encoding system and the decoding system of FIGS. 1 and 2 and an update process thereof.
FIG. 10 is a flowchart showing an encoding operation in the encoding system of FIG. 1;
11 is a diagram for explaining the handling of peripheral pixels for a two-dimensional color order table in the encoding operation of the encoding system of FIG. 1, and FIG. 11A illustrates the case of encoding the first encoding target pixel. (B) is a figure for demonstrating the case where the encoding object pixel in the 1st line is encoded.
12 is a diagram for explaining the handling of peripheral pixels for the two-dimensional color order table in the encoding operation of the encoding system of FIG. 1, and (A) is a diagram when the encoding target pixel comes to the beginning of the second line. FIG. 4B is a diagram for explaining encoding, and FIG. 5B is a diagram for explaining encoding when the encoding target pixel comes to the end of the second line, and FIG. It is a figure for demonstrating encoding when it comes to the first part.
FIG. 13 is a diagram for explaining the handling of peripheral pixels for a two-dimensional color order table and reference pixels for a Markov model in the encoding operation of the encoding system of FIG. 1, and a narrow strip is the same as a normal strip; It is a figure which shows the case where it can handle with thought.
14 is a diagram for explaining the handling of peripheral pixels for a two-dimensional color order table and reference pixels for a Markov model in the encoding operation of the encoding system of FIG. 1, and it is necessary to perform exception processing different from a normal strip; (A) shows a case where the horizontal width is 3 pixels, (B) shows a case where it has 2 pixels, and (C) shows a case where it has 1 pixel.
FIG. 15 is a flowchart showing a decoding operation in the decoding system of FIG. 2;
FIG. 16 is a block diagram of a conventional multi-color image encoding system and decoding system.
FIG. 17 is an explanatory diagram of reference pixel data with respect to conventional encoding target pixel data.
FIG. 18 is a diagram showing a conventional parameter table.
FIG. 19 is a block diagram of a conventional multi-color image encoding system and decoding system having a state degeneration unit.
FIG. 20 is a diagram illustrating an example of a conventional degeneration table.
FIG. 21 is an explanatory diagram of a conventional arithmetic code type entropy encoder and entropy decoder;
FIG. 22 is a diagram for explaining an index of a conventional multi-color image.
FIGS. 23A and 23B are explanatory diagrams showing the principle of creating a conventional combined color order table, where FIG. 23A shows the arrangement relationship of each pixel, FIG. 23B shows the color symbol of each pixel, and FIG. It is a figure which shows the performed color order table (latest appearance table).
[Explanation of symbols]
1 Coding system
3 Decryption system
5 Strip generation means
7 Strip synthesis method
10 line buffer (peripheral pixel generation means)
11 Markov model generator (degeneration means)
12 Entropy encoder (entropy encoding means)
13 Predictor (prediction means)
14 color order generator
15 Discriminator (also serves as escape code generation means)
16 Two-dimensional color order generation unit
17 One-dimensional color order generator
18 Prediction table composition part
19 Priority switching unit
20 2D color order table
21 Table update section
22 One-dimensional color order table
23 Combined color order table
30 line buffer
31 Markov model generator (degeneration means)
32 Entropy decoder (entropy decoding means)
33 Predictor
100A color pixel data to be encoded
100B Decoded color pixel data
120 color ranking data
200 encoded data

Claims (4)

In a multi-color image encoding device that encodes input target color pixel data into encoded data and outputs the encoded data,
A strip generation unit that divides a target image to be encoded into strips having a limited number of pixels and outputs pixel data corresponding to the strips;
Based on the pixel data corresponding to the strip, the peripheral pixel data for the two-dimensional peripheral pixel corresponding to the pixel data and the reference pixel data for the peripheral four reference pixels excluding the pixel immediately before the pixel to be encoded are generated. And line buffer means for outputting the peripheral pixel data and the reference pixel data,
Create a color order table based on the surrounding pixel data, create a Markov model context based on the reference pixel data,
An encoding apparatus that sequentially encodes the encoding target pixels located in the strips based on the color order table and the context. In the encoding method of the multi-color image which encodes the object color pixel data to be input into encoded data and outputs the encoded color pixel data,
A strip generation step of dividing the target image to be encoded into strips having a limited number of pixels and outputting pixel data corresponding to the strips;
Based on the pixel data corresponding to the strip, the peripheral pixel data for the two-dimensional peripheral pixel corresponding to the pixel data and the reference pixel data for the peripheral four reference pixels excluding the pixel immediately before the pixel to be encoded are generated. And a line buffer step for outputting the peripheral pixel data and the reference pixel data,
A color rank table is created based on the peripheral pixel data, a Markov model context is created based on the reference pixel data, and the code located in each strip based on the color rank table and the context An encoding method characterized by sequentially encoding pixels to be encoded. In a multi-color image decoding apparatus that decodes input target encoded data into decoded data and outputs the decoded data,
Strip synthesis means for decoding a target image to be decoded by dividing the target image into strips having a horizontal width of a limited number of pixels;
Line buffer means for generating two-dimensional peripheral pixel data and reference pixel data corresponding to the pixel data based on the pixel data corresponding to the strip and outputting the peripheral pixel data and the reference pixel data; ,
Create a color order table based on the surrounding pixel data, create a Markov model context based on the decoded data of the surrounding 4 pixels excluding the previous pixel ,
A decoding apparatus that sequentially decodes the target images to be decoded in a strip shape based on the color order table and the context. In a decoding method of a multi-color image in which input target encoded data is decoded into decoded data and output,
A strip synthesis step of decoding a target image to be decoded by dividing the target image into strips having a limited number of pixel widths; and
A line buffer step of generating two-dimensional peripheral pixel data and reference pixel data corresponding to the pixel data based on the pixel data corresponding to the strip, and outputting the peripheral pixel data and the reference pixel data; ,
Create a color order table based on the surrounding pixel data, create a Markov model context based on the decoded data of the surrounding 4 pixels excluding the previous pixel ,
A decoding method comprising: sequentially decoding the target images to be decoded in a strip shape based on the color order table and the context.

Images