JP3705771B2 - JBIG decoding method - Google Patents

Description

[0001]
[Technical field to which the invention belongs]
The present invention relates to a binary image decoding method, and more particularly, to a binary image decoding method that speeds up the decoding process by JBIG (joint bi-level image group).
[0002]
[Prior art]
Conventionally, when decoding a monochrome binary image such as a facsimile, various processing methods are known when one line is all white.
[0003]
For example, in the facsimile decoding circuit disclosed in Japanese Patent Laid-Open No. 1-198868, when the decoding result is all white for one line, image data is not transferred to the line memory, and this is controlled by a control signal. By notifying, the same result as when one-line image data is transferred is obtained.
[0004]
Also, in the decompression processing apparatus for compressed code data disclosed in Japanese Patent Application Laid-Open No. 1-222379, when data corresponding to all white for one line is input during decompression processing, the image data of that line is stored. Even if the image data is not generated / transferred, the same effect as that of sequentially generating / transferring the image data is obtained simply by increasing the address counter.
[0005]
In the facsimile decoding apparatus disclosed in Japanese Patent Application Laid-Open No. Hei 2-213278, when one reference line is used for two-dimensional decoding, the image data of the reference line is not read. ing. As a result, the two-dimensional decoding can be speeded up.
[0006]
In the facsimile encoding apparatus and decoding apparatus disclosed in Japanese Patent Laid-Open No. 3-228470, when there is a line in which all pixels are white, a 1-bit all white code is generated. Then, a 1-bit identification code for displaying the all white code is also added. That is, one line all white can be expressed by 2 bits. Therefore, encoding / decoding is executed at each stage at a higher speed than MH (modified Huffman) code and MR (modified READ) code.
[0007]
In the facsimile apparatus disclosed in Japanese Patent Publication No. 4-23991, the run length of binary data is compressed using a modified read (MR) code or a modified modified read (MMR) code. Is determined whether there is a change point in the reference line, that is, the previous line, and the determination result and the code of the line to be decoded are used to determine whether all white lines are continuous. Yes. If this continues, the line is not decoded, and the printing from the line memory for printing and the printing recording operation are not performed.
[0008]
Furthermore, JBIG (Joint bi-label Image coding group, ITU-T.82 / ISO / IEC11544) is one of data compression / decompression techniques. In JBIG encoding, the level of a pixel to be encoded is estimated from its surrounding area. Specifically, the peripheral area of the target pixel, that is, the model template is made one-dimensional to obtain the context. Then, the appearance probability of the level of the target pixel corresponding to the context is estimated by learning. Then, based on the learning result, the level of the target pixel is arithmetically encoded (AAE) and expressed as a binary decimal number of 0 or more and less than 1. In the case of JBIG decoding, the reverse procedure is performed.
[0009]
In JBIG encoding, since prediction learning is performed on whether the target pixel is white or black, encoding is performed only when the prediction is lost. Performing encoding / decoding using this JBIG is disclosed in, for example, Japanese Patent No. 2919384.
[0010]
Here, the JBIG model template (Model Template: MT) process will be described.
[0011]
FIG. 9 shows a memory block for storing image data. The memory block stores data of the decoding target line PIX, the previous reference line H1, and the second previous reference line H2.
[0012]
FIG. 10 shows a model template in the memory block. The model template is a peripheral pixel pattern of the target pixel to be decoded.
[0013]
There are two model templates: a 3-line model template and a 2-line model template. The shape of MT is determined in advance.
[0014]
However, when encoding an image having a strong correlation with a certain period, the position of the second pixel of the 3-line MT or the position of the fourth pixel of the 2-line MT may be moved to, for example, 127 pixels. .
[0015]
As shown in FIG. 10, the context is a one-dimensional arrangement of the ninth pixel to the zeroth pixel of MT in this order.
[0016]
11A and 11B show specific examples of contexts. The context of FIG. 11A does not include black, and this context is represented by “000H”. On the other hand, in the context of FIG. 11B, the ninth pixel and the fifth pixel are black, and this context is represented by “220H”.
[0017]
FIG. 12 shows a specific example of a context table, that is, a learning table. This table is a table for storing pixel values predicted from context values obtained from a model template for 10 pixels and their state numbers. The current state (ST) is set to “0”. Then, it is assumed that the target pixel is predicted to be white “0” in the current learning table (context “220H”). If the prediction is correct, the learning table is left as it is. On the other hand, when the prediction is wrong, the prediction value is changed to black “1” and ST is set to “1”.
[0018]
FIG. 13 shows a probability estimation table. This table is composed of the LPS region width LSZ, the next state transition numbers NLPS and NMPS indicated by the context table, that is, the learning table, and the predicted value inversion SWITCH. Here, in the symbols to be encoded / decoded, the dominant symbol MPS is used as the symbol that appears more frequently, and the inferior symbol LPS as the symbol that appears less. In addition, the state transition destination at the time of LPS, that is, unexpected, is NLPS.
In addition, the state transition destination at the time of MPS, that is, the expected hit is NMPS. Further, if SWITCH = 1 at the time of LPS, that is, when it is not expected, the predicted value is inverted.
[0019]
According to this probability estimation table, if the prediction is lost when the state number ST = 0, the state is shifted to the state number NLPS = 1 and the predicted value must be inverted. Therefore, the prediction value of the context table shown in FIG. 12, that is, the learning table is changed from white to black, and the state number is changed from “0” to “1”.
[0020]
FIG. 14 is a block diagram of the JBIG data decoding apparatus. This apparatus includes a CPU (Central Processing Arithmetic Unit) 100 that performs arithmetic processing, data processing, control processing of each unit, an image data memory 400 that stores a binary image of a document read at the time of transmission, and at the time of reception. A code data memory 500 that stores received data as codes, a ROM 200 in which a decoding program 201 by JBIG and a probability estimation table 202 for converting code data into image data are arranged and stored, and a pixel to be decoded is white or black A RAM 300 in which a learning table 301 that is referred to in order to increase the probability of being predicted is placed and stored, and a FIFO memory 600 that manages image data that is the result of conversion is a bus. 700 is connected.
[0021]
The CPU 100 includes a register 101 that holds image data read in units of blocks from a decoding target line, a register 102 that holds image data read out in units of blocks from a line immediately preceding the decoding target line, and a decoding target A register 103 that holds image data read in units of blocks from the line two lines before the line, a register 104 that holds prediction results, a register 105 that holds prediction values and state values read from the learning table 301, and a probability It has a register 106 that holds an out-of-range area width read from the estimation table 202 and a register 107 that holds the contents of the context. The CPU 100 executes JBIG decryption processing by executing a JBIG decryption program 201.
[0022]
FIG. 15 is a flowchart for explaining the JBIG decoding method.
[0023]
First, it is determined whether or not to perform typical prediction (TP), which is an option of JBIG (S21). Here, the typical prediction TP is a process for reducing the number of target pixels to be encoded / decoded, that is, the number of target pixels. In the case of encoding processing, when the encoding target line PIX is the same as the previous line H1, encoding is not performed. The fact that the PIX image data and the H1 image data are the same is encoded as a pseudo pixel using 1 bit. On the other hand, in the case of decoding processing, if the decoding target line PIX is the same as the previous line H1, decoding is not performed and the image data of H1 is output (copied).
[0024]
When the typical prediction as described above is not performed, a one-line decoding process is performed (S210). Subsequent to S210, it is determined whether processing for the number of document lines has been completed (S27). If not completed, the process returns to 1-line decoding processing (S210) and the decoding processing is repeated.
[0025]
On the other hand, when typical prediction is performed (S21), it is determined whether or not the two lines (the decoding target line and the preceding line of the decoding target line) are the same (S29). If the two lines are not the same, a one-line decoding process (S210) is performed. On the other hand, if the two lines are the same, the decoding target line is not subject to decoding. Next, as in the case where typical prediction is not performed (S21), it is determined whether processing for the number of document lines has been completed (S27). If not, the process returns to line matching confirmation processing (S29). Repeat the decryption process.
[0026]
The above process is repeated for the number of document lines, for example, 2376 lines in the case of an A4 size document, so that the decoding process is performed.
[0027]
FIG. 16 is a flowchart for explaining the one-line decoding process in S210 of FIG.
[0028]
First, the three image data blocks of the line H2 immediately before the decoding target line, the line H1 immediately before the decoding target line, and the decoding target line PIX are read in block units (S51).
[0029]
Next, it is determined whether or not the three image data blocks are all white (S52).
[0030]
If the determination result is all white, the one-pixel decoding process (S55) is repeated until the block process ends. When the process for each block is completed, the process returns to the beginning of the one-line decoding process to determine whether the decoding process for one line is completed (S50). The process ends. On the other hand, if the line is not finished as a result of the discrimination, a series of decoding processes in units of blocks from the image data reading process (S51) is continued.
[0031]
On the other hand, if the determination result is not all white in S52, a context in which the peripheral pixels (model template) of the decoding target pixel is made one-dimensional is created (S53) until the block processing is completed (S56). A decoding process is performed (S54), and a series of processes for shifting the image data of three blocks by 1 bit each (S57) is repeated. When the process for each block is completed, the process returns to the beginning of the one-line decoding process to determine whether the decoding process for one line is completed (S50). The process ends.
[0032]
On the other hand, if the determination result is not the end of line in S50, a series of decoding processes in units of blocks from the image data reading process (S51) is continued.
[0033]
The decoding process is performed by repeating the above process for the number of document lines.
[0034]
FIG. 17 is a flowchart for explaining the one-pixel decoding process in S54 of FIG. This one-pixel decoding process in S54 is different from the one-pixel decoding process in S55.
[0035]
As shown in FIG. 17, first, a prediction value and a state value are read from the learning table 301 using a context obtained by making the peripheral pixels of the decoding target pixel one-dimensionally as an index (S120).
[0036]
Next, a region width out of the probability estimation table 202 is read using the state value as an index (S31).
[0037]
Further, the deviated region width is subtracted from the region width indicating the probability that a black and white permutation appears, and updated to a new region width (S40).
[0038]
Next, it is determined whether the decoding target pixel matches the predicted value (S121). When it is determined that the pixel to be decoded does not match the predicted value, the region width indicating the probability of divergence is changed, and a prediction deviance process for writing the updated state value in the learning table 301 is performed (S122). Normalization is performed to make the width larger than the outlying region width (S43).
[0039]
On the other hand, if it is determined in S121 that the decoding target pixel matches the predicted value, it is determined whether normalization is required to make the region width larger than the outlying region width (S41). When normalization is required in the determination result, the region width indicating the probability of winning is changed, and the prediction hit process (S42) and the normalization process (S43) for writing the updated state value in the learning table 301 are performed. A series of one-pixel decoding processes are terminated.
[0040]
FIG. 18 is a flowchart for explaining the one-pixel decoding process in S55 of FIG.
[0041]
As shown in FIG. 18, first, a prediction value and a state value are read out from the learning table 301 by using, as an index, a context (0) obtained by making the peripheral pixels of the decoding target pixel one-dimensional (S30).
[0042]
Next, a region width out of the probability estimation table 202 is read using the state value as an index (S31). Further, the deviated region width is subtracted from the region width indicating the probability that a black and white permutation appears, and updated to a new region width (S40).
[0043]
Next, it is determined whether or not the predicted value of the pixel to be decoded is white (S130). When it is determined that the predicted value of the decoding target pixel is not white, the region width indicating the probability of divergence is changed, and an unpredictive process for writing the updated state value in the learning table 301 is performed (S122). Normalization is performed so as to be larger than the outlying region width (S43).
On the other hand, if it is determined in S130 that the predicted value of the decoding target pixel is white, it is determined whether normalization is required to make the region width larger than the outlying region width (S41). When normalization is required in the determination result, the region width indicating the probability of winning is changed, and the prediction hit process (S42) and the normalization process (S43) for writing the updated state value in the learning table 301 are performed. A series of one-pixel decoding processes are terminated.
[0044]
On the other hand, if normalization is not required in S41, it is determined whether or not the processing for each block has been completed (S131). If not, the predicted value, state value, and probability estimation table from the learning table 301 are determined. Without reading out the outlying area from 202, a series of processes for updating only the area width (S40) is repeated.
[0045]
On the other hand, if the determination result is block end in S131, the one-pixel decoding process is ended.
[0046]
Summarizing the features of the prior art described above, if the image data read in units of blocks is all white, the context creation process is not necessary, and if the predicted value is white and the normalization process is not necessary, the learning table ( (Predicted value, state value) Further, since it is not necessary to read data from the probability estimation table 202 (outlier area width), the decoding processing speed can be improved.
[0047]
[Problems to be solved by the invention]
However, the above-described conventional technology lacks processing performance for real-time communication when typical prediction is not performed. The first reason is that image data block read processing for context creation is frequently performed on a block-by-block basis for the three lines of the decoding target line, the previous line, and the previous line. Because it is done. The second reason is that an all white discrimination process is performed for each block unit and a prediction value discrimination process is performed for each one-pixel decoding.
[0048]
Accordingly, an object of the present invention is to further improve the JBIG decoding processing speed by reducing the image data read processing, block discrimination processing, and prediction value discrimination processing for three lines.
[0049]
Another object of the present invention is to reduce the number of times the image data, the learning table, and the probability estimation table are read from each memory as the JBIG decoding processing speed is improved, thereby reducing the power consumption of the JBIG decoding device. Yes.
[0050]
[Means for Solving the Problems]
A JBIG decoding method according to the present invention is a JBIG decoding method for decoding a decoding target pixel on the decoding target line using a decoding target line and one or two reference lines preceding the decoding target line. When the reference line is all white, it is determined whether or not the predicted value of the context whose content is all 0 is white. When the predicted value is white, the first one-line decoding process dedicated to all white On the other hand, when the reference line is not all white or when the predicted value of the context is not white, a second one-line decoding process is performed in which the peripheral pixels of the decoding target pixel are read after checking the all white flag, In the second one-line decoding process, when the reference line is all white, the image data reading process from the reference line is not performed.
[0051]
The JBIG decoding method according to the present invention is a JBIG decoding method for decoding a decoding target pixel on the decoding target line using a decoding target line and one or two reference lines preceding the decoding target line. In the case where the typical prediction is not performed and the reference line is all white, it is determined whether or not the prediction value of the context whose contents are all 0 is white. On the other hand, when the reference line is not all white or when the predicted value of the context is not white, the reading process of the peripheral pixels of the decoding target pixel is performed after the all white flag check. In the second one-line decoding process, when the reference line is all white, the image data is read from the reference line. In the case of typical prediction, it is determined whether or not the decoding target line is the same as the previous reference line, and if not, the periphery of the decoding target pixel A third one-line decoding process is performed in which a pixel reading process is performed every block, and if the same, the decoding target line is not set as a decoding target.
[0052]
The JBIG decoding method according to the present invention is a JBIG decoding method for decoding a decoding target pixel on the decoding target line using a decoding target line and one or two reference lines preceding the decoding target line. In the case of typical prediction, it is determined whether or not the decoding target line and the preceding reference line are the same, and if they are the same, an all white flag setting process is performed, but not the same. Sometimes, when the reference line is all white, it is determined whether the predicted value of the context whose content is all 0 is white, and when the predicted value is white, the first one-line decoding dedicated to all white On the other hand, when the reference line is not all white or when the predicted value of the context is not white, the process of reading the peripheral pixels of the decoding target pixel is performed after checking the all white flag. The second one-line decoding process is performed, and in the second one-line decoding process, when the reference line is all white, the image data reading process from the reference line is not performed. In the case where typical prediction is not performed, a third one-line decoding process is performed in which the reading process of the peripheral pixels of the decoding target pixel is performed every block.
[0053]
A JBIG decoding program according to the present invention is a JBIG decoding method for decoding a decoding target pixel on the decoding target line using a decoding target line and one or two reference lines preceding the decoding target line. When the reference line is all white, it is determined whether or not the predicted value of the context whose content is all 0 is white. When the predicted value is white, the first one-line decoding process dedicated to all white On the other hand, when the reference line is not all white or when the predicted value of the context is not white, a second one-line decoding process is performed in which the peripheral pixels of the decoding target pixel are read after checking the all white flag, In the second one-line decoding process, when the reference line is all white, the image data reading process from the reference line is not performed. .
[0054]
The JBIG decoding program according to the present invention is a JBIG decoding program for decoding a decoding target pixel on the decoding target line using a decoding target line and one or two reference lines preceding the decoding target line. In the case where typical prediction is not performed, if the reference line is all white, it is determined whether or not the prediction value of the context whose contents are all 0 is white. On the other hand, when the reference line is not all white or when the predicted value of the context is not white, the reading process of the peripheral pixels of the decoding target pixel is performed after the all white flag check. In the second one-line decoding process, if the reference line is all white, the image from the reference line is In the case of typical prediction, it is determined whether or not the decoding target line is the same as the previous reference line, and if not, the decoding target A third one-line decoding process is performed in which a pixel neighboring pixel reading process is performed every block, and if the same, the decoding target line is not set as a decoding target.
[0055]
The JBIG decoding program according to the present invention is a JBIG decoding program for decoding a decoding target pixel on the decoding target line using a decoding target line and one or two reference lines preceding the decoding target line. In the case of typical prediction, it is determined whether or not the decoding target line and the preceding reference line are the same, and if they are the same, an all white flag setting process is performed, but not the same. Sometimes, when the reference line is all white, it is determined whether the predicted value of the context whose content is all 0 is white, and when the predicted value is white, the first one-line decoding dedicated to all white On the other hand, when the reference line is not all white or when the predicted value of the context is not white, the reading processing of the peripheral pixels of the pixel to be decoded is all white. A second one-line decoding process performed after checking the image, and in the second one-line decoding process, when the reference line is all white, the image data reading process from the reference line is not performed. In a case where typical prediction is not performed, a third one-line decoding process is performed in which a process of reading pixels around a decoding target pixel is performed every block.
[0060]
DETAILED DESCRIPTION OF THE INVENTION
Embodiments of the present invention will be described below with reference to the drawings.
[0061]
FIG. 1 is a block diagram of a decoding apparatus used in the JBIG decoding method of the present invention. This apparatus includes a CPU (Central Processing Arithmetic Unit) 100 that performs arithmetic processing, data processing, control processing of each unit, a code data memory 500 that stores code data received at the time of reception, and image data that is a result of conversion. Image data memory 400, JBIG decoding program 201A and ROM 200 in which probability estimation table 202 for converting code data into image data is arranged and stored, and whether the decoding target pixel is white or black A bus 300 includes a RAM 300 in which a learning table 301 referred to in order to increase the probability of hitting, and a FIFO (first in first out) managed FIFO memory 600 that temporarily stores code data or image data. It is connected.
[0062]
The CPU 100 also includes a register 101 that holds image data read in units of blocks from the decoding target line, a register 102 that holds image data read out in units of blocks from the previous line of the decoding target line, and a decoding A register 103 that stores image data read in units of blocks from a line immediately before the conversion target line, a register 104 that stores prediction results, and a register 105 that stores prediction values and state values read from the learning table 301. The register 106 that holds the out-of-range area width read from the probability estimation table 202, the register 107 that holds the contents of the context, and whether the line immediately preceding the decoding target line (H2 line) is all white. The H2 flag and the line before the decoding target line (H1 line) are over. H1 flag indicating whether the white and decoded line has a register 108 for holding the PIX flag that indicates whether all white.
[0063]
[First Embodiment]
The first embodiment is a case where the model template is three lines. Processing of the JBIG program unit 201A in the ROM 200 of FIG. 1 will be described with reference to the flowcharts of FIGS.
[0064]
The overall operation will be described with reference to FIG.
[0065]
First, at the beginning of the decoding process for one page, the all white flag (H2 flag: 2 lines before, H1 flag: 1 line before, PIX flag: 3 flags of the decoding target line) is initialized ( S20).
When the decoding target line is at the head of the page, there are no two lines before and one line before, but actually there is a virtual white line.
[0066]
Next, it is determined whether or not typical prediction is performed (S21). If typical prediction is not performed, an all white flag resetting process (S22) is performed. Specifically, the all white flag is substituted for H2 one line before from H1 one line before, and the all white flag is substituted for H1 one line before PIX of the decoding target line (S22).
[0067]
Next, it is determined whether the reference line (H2 line, H1 line) is all white or not all white (S23).
[0068]
If the image data of the reference line is all white by the determination process, it is determined whether or not the predicted value of the context (0) is white (S24). 1-line decoding process 1 is performed (S25).
[0069]
On the other hand, if it is determined in S23 that the reference line is not all white, or if the predicted value is not white in S24, a second one-line decoding process is performed (S26).
[0070]
Next, it is determined whether or not the processing for the entire number of document lines has been completed (S27). If the processing has not been completed, S29 (line matching confirmation processing) and S22 (all white) are performed according to the presence or absence of typical prediction settings. Returning to the flag resetting process, the decoding process is repeated.
[0071]
When typical prediction is performed, it is determined whether or not the two lines (the decoding target line and the preceding line of the decoding target line) are the same after S27 (S29). If the determination results are not the same, the conventional one-line decoding process (S210) is executed, and if it is determined that they are the same, the decoding target line PIX is not set as the decoding target.
[0072]
Next, following S29 or S210, it is determined whether or not the processing for the number of document lines has been completed (S27). If not completed, the processing returns to the line matching confirmation processing (S29) and the decoding processing is repeated.
[0073]
On the other hand, when the typical prediction is not performed, the process returns to the all white flag resetting process (S22) following S27.
[0074]
The decoding process is performed by repeating the above process for the number of document lines.
[0075]
Next, referring to FIG. 3, a first one-line decoding process dedicated to all white when the reference lines (H2 line, H1 line) are all white and the predicted value of context (0) is white (FIG. 2). Will be described.
[0076]
First, a prediction value and a state value are read from the learning table 301 by using, as an index, a context (0) obtained by making the peripheral pixels of the decoding target pixel one-dimensional (S30).
[0077]
Next, using the state value as an index, an outlier region width is read from the probability estimation table 202 (S31), and a one-pixel decoding process (S32) is performed.
[0078]
Next, the decoding result of the one-pixel decoding process is determined (S32). If the decoding result is white, the one-pixel decoding process is repeated until the end of the line (S34), and finally the all white flag PIX Is set (S35), and the series of processes is terminated.
[0079]
If it is determined in the decoding result that the decoding result is not white, the all white flag PIX is reset (S36), and the process branches to the second one-line decoding process (branch A).
[0080]
If the line end determination (S34) is not the end of the line, depending on the presence or absence of the normalization process during the one-pixel decoding process (S37), S30 (predicted value and state value reading process), S32, respectively. Returning to (one pixel decoding process), the decoding process is repeated.
[0081]
Next, the one-pixel decoding process S32 of FIG. 3 will be described with reference to FIG.
[0082]
First, the outlying area width is subtracted from the area width indicating the probability that a black and white permutation appears, and updated to a new area width (S40).
[0083]
Next, it is determined whether the region width is reduced by the region width updating process, and normalization is required to make the region width larger than the outlying region width (S41).
[0084]
If normalization is required based on the determination result, a prediction hit process (S42) in which the region width indicating the hit probability is changed and the updated state value is written in the learning table 301 is performed, and a normalization process is performed (S43). If normalization is not necessary, the one-pixel decoding process is terminated without doing anything.
[0085]
In the first one-line decoding process of S25 of FIG. 2, since it is known that the reference line is all white and the predicted value of the context (0) is white, the one-pixel decoding result is not white. In other words, the value of the model template is always white (0) and the context is always “0” until it becomes black. For this reason, the reading process, the block determination process, and the prediction value determination process of the peripheral pixels (model template) around the decoding target pixel, which have been performed for each block unit, are unnecessary.
[0086]
Next, the second one-line decoding process in S26 of FIG. 2 will be described with reference to FIG. As described above, the second one-line decoding process is performed when the reference line is not all white in the all-white discrimination process (S23) or when the decoding result is obtained during the first one-line decoding process. If it is not white, it is executed after branching (branch A).
[0087]
Basically, the same processing as the conventional one is performed, but the difference is that the peripheral pixel reading processing (S51 in FIG. 16 in the prior art) in the present invention is performed after the all white flag check (S520) in the present invention. Is to be performed (S530).
[0088]
Another difference is that after decoding one line, it is determined whether the decoded line is all white (S58). If it is all white, the all white flag PIX is set (S59), and not all white. In this case, a series of processes for resetting the all white flag PIX (S510) is added.
[0089]
Specifically, during the reading process of the image data, the state of the all white flag is checked (S520), and the reading process of the image data from the reference line in which the all white flags H2 and H1 are set is not performed ( S530). If none of the flags is set in the all white flag check, the image data reading process is performed in the same manner as before (S530).
[0090]
By performing the above processing, the decoding processing speed can be improved.
[0091]
It is obvious that the present invention is particularly effective for an image with a tendency to white eyes, such as a document image often handled by a facsimile.
[0092]
[Second Embodiment]
The second embodiment is a case where the model template is two lines.
[0093]
Referring to FIG. 6, in the second embodiment, the all white flag is initialized in all white flag initialization (S60) and all white flag reset processing (S61) in the first embodiment described in FIG. The difference is that H2 is not used. This is because when the model template is two lines, the reference line is only the line H1 immediately preceding the decoding target line. Other processes are the same as those in the first embodiment. By performing the above processing, the decoding processing speed can be improved as in the first embodiment.
[0094]
[Third Embodiment]
The third embodiment is a case where the model template is three lines. As shown in FIG. 7, when typical prediction is performed, the reference line all white discrimination process (S23), the context (0) prediction value discrimination process (S24) described in the first embodiment, and all white The first one-line decoding process (S25) and the second one-line decoding process (S26) are applied.
[0095]
The basic process is the same as that of the first embodiment, except that an all white flag setting process (S70) is added. The reason for this is that in the case of typical prediction, the two lines (the line to be decoded and the line preceding the line to be decoded) are identical in the line matching confirmation process (S29) for determining whether or not the two lines are the same. This is because the one-line decoding process is not performed when it is determined that they are the same. When it is determined that the two lines are the same, the process of actually copying the previous line is performed.
[0096]
By performing the above processing, the decoding processing speed can be improved as in the first embodiment.
[0097]
[Fourth Embodiment]
The fourth embodiment is a case where the model template is two lines. As shown in FIG. 8, the all white flag initialization (S60) and the all white flag resetting process (S61) of the third embodiment are different in that the all white flag H2 is not used. This is because when the model template is two lines, the reference line is only the line H1 immediately preceding the decoding target line. Other processes are the same as those in the third embodiment. By performing the above processing, the decoding processing speed can be improved as in the first embodiment.
[0098]
【The invention's effect】
According to the present invention described above, the decoding processing speed can be improved (a speed increase of several tens of percent can be expected as compared with the prior art). The reason for this is that processing for determining whether the reference line is all white and processing for determining whether the predicted value of the context (0) is white when the determination result is all white are all white lines. This is because branching to the first one-line decoding process can reduce the image data reading process, block discrimination process, and predicted value discrimination process for three lines. The second reason is that in the first one-line decoding process for the all white lines, in the second one-line decoding process that is branched and executed when the decoding result is not white (black), This is because it is known that the line is all white, so that it is not necessary to read out image data from the reference line.
[0099]
Further, according to the present invention, the power consumption of the system in the decoding process can be reduced. The reason is that the number of times of reading the image data, the learning table, and the probability estimation table from each memory can be reduced.
[Brief description of the drawings]
FIG. 1 is a block diagram of a decoding apparatus used in a JBIG decoding method of the present invention.
FIG. 2 is a flowchart for explaining a decoding method according to the first embodiment of the present invention;
FIG. 3 is a flowchart for explaining a first one-line decoding process in S25 of FIG. 2;
FIG. 4 is a flowchart for explaining the one-pixel decoding process in S32 of FIG. 3;
FIG. 5 is a flowchart for explaining a second one-line decoding process in S26 of FIG. 2;
FIG. 6 is a flowchart for explaining a decoding method according to the second embodiment;
FIG. 7 is a flowchart for explaining a decoding method according to the third embodiment;
FIG. 8 is a flowchart for explaining a decoding method according to the fourth embodiment;
FIG. 9 is a diagram showing a memory block for storing image data.
FIG. 10 is a diagram showing a model template
FIG. 11A is a diagram showing an example of a context
FIG. 11B is a diagram showing another example of the context
FIG. 12 is a diagram showing an example of a context table, that is, a learning table.
FIG. 13 shows a probability estimation table.
FIG. 14 is a block diagram of a conventional JBIG decoding apparatus.
FIG. 15 is a flowchart for explaining a conventional JBIG decoding method;
FIG. 16 is a flowchart for explaining the one-line decoding process in S210 of FIG. 15;
FIG. 17 is a flowchart for explaining the one-pixel decoding process in S54 of FIG. 16;
FIG. 18 is a flowchart for explaining the one-pixel decoding process in S55 of FIG.
[Explanation of symbols]
100 CPU
200 JBIG program and probability estimation table storage ROM
300 Learning table storage RAM
400 Image data storage RAM
500 Code data storage RAM
600 RAM (FIFO memory)
700 buses

Claims (6)

A JBIG decoding method for decoding a decoding target pixel on the decoding target line using a decoding target line and one or two reference lines preceding the decoding target line,
When the reference line is all white, it is determined whether or not the predicted value of the context whose contents are all 0 is white. When the predicted value is white, the first one-line decoding process dedicated to all white is performed. On the other hand, when the reference line is not all white or when the predicted value of the context is not white, a second one-line decoding process is performed in which the peripheral pixel of the decoding target pixel is read after checking the all white flag, In the one-line decoding process of No. 2, when the reference line is all white, the image data reading process from the reference line is not performed. A JBIG decoding method for decoding a decoding target pixel on the decoding target line using a decoding target line and one or two reference lines preceding the decoding target line,
In a case where typical prediction is not performed, if the reference line is all white, it is determined whether or not the prediction value of the context whose contents are all 0 is white. 1 line decoding processing of 1 is performed, while when the reference line is not all white or when the predicted value of the context is not white, the second 1 line is read after the all white flag is read out. In the second one-line decoding process, when the reference line is all white, the image data read-out process from the reference line is not performed.
When typical prediction is performed, it is determined whether or not the decoding target line is the same as the reference line immediately before the decoding target line. If not, the process of reading out the surrounding pixels of the decoding target pixel is performed every block. Execute a third one-line decoding process;
A JBIG decoding method characterized in that the decoding target line is not set as a decoding target when they are the same. A JBIG decoding method for decoding a decoding target pixel on the decoding target line using a decoding target line and one or two reference lines preceding the decoding target line,
In a typical prediction, it is determined whether or not the decoding target line and the previous reference line are the same,
If they are the same, the all white flag setting process is performed,
If they are not the same, if the reference line is all white, it is determined whether or not the predicted value of the context whose contents are all 0 is white. If the predicted value is white, the first one line dedicated to all white While performing the decryption process
When the reference line is not all white or when the predicted value of the context is not white, a second one-line decoding process is performed in which the peripheral pixel of the decoding target pixel is read after checking the all white flag, and the second one In the line decoding process, when the reference line is all white, the image data reading process from the reference line is not performed,
A JBIG decoding method characterized by performing a third one-line decoding process in which a reading process of peripheral pixels of a decoding target pixel is performed every block when typical prediction is not performed. A JBIG decoding program for decoding a decoding target pixel on the decoding target line using a decoding target line and one or two reference lines preceding the decoding target line,
When the reference line is all white, it is determined whether or not the predicted value of the context whose contents are all 0 is white. When the predicted value is white, the first one-line decoding process dedicated to all white is performed. On the other hand, when the reference line is not all white or when the predicted value of the context is not white, a second one-line decoding process is performed in which the peripheral pixel of the decoding target pixel is read after checking the all white flag, In the one-line decoding process of No. 2, when the reference line is all white, the image data reading process from the reference line is not performed. A JBIG decoding program for decoding a decoding target pixel on the decoding target line using a decoding target line and one or two reference lines preceding the decoding target line,
In a case where typical prediction is not performed, if the reference line is all white, it is determined whether or not the prediction value of the context whose contents are all 0 is white. 1 line decoding processing of 1 is performed, while when the reference line is not all white or when the predicted value of the context is not white, the second 1 line is read after the all white flag is read out. In the second one-line decoding process, when the reference line is all white, the image data read-out process from the reference line is not performed.
When typical prediction is performed, it is determined whether or not the decoding target line is the same as the reference line immediately before the decoding target line. If not, the process of reading out the surrounding pixels of the decoding target pixel is performed every block. Execute a third one-line decoding process;
A JBIG decoding program characterized in that when the same, the decoding target line is not set as a decoding target. A JBIG decoding program for decoding a decoding target pixel on the decoding target line using a decoding target line and one or two reference lines preceding the decoding target line,
In a typical prediction, it is determined whether or not the decoding target line and the previous reference line are the same,
If they are the same, the all white flag setting process is performed,
If they are not the same, if the reference line is all white, it is determined whether or not the predicted value of the context whose contents are all 0 is white. If the predicted value is white, the first one line dedicated to all white While the decoding process is performed, when the reference line is not all white or when the predicted value of the context is not white, the second one-line decoding process is performed after the all white flag is checked when the peripheral pixel of the decoding target pixel is read. Done
In the second one-line decoding process, when the reference line is all white, the image data reading process from the reference line is not performed,
A JBIG decoding program characterized by performing a third one-line decoding process in which a process of reading out neighboring pixels of a decoding target pixel is performed every block when typical prediction is not performed.

Images