JPH06113142A - Method and device for processing picture - Google Patents

Abstract

PURPOSE:To provide method/device by which a code data amount can be controlled to a target data amount. CONSTITUTION:In the method/device by which the code data amount can be controlled to the target data amount, a quantized quantizing coefficient is encoded in a Huffman encoding part 14, code data and a code length are stored in a buffer memory 15, the code length is also inputted to a code length adder 17 and the added result is compared with the value of a threshold table 19 in a comparator 18. A stage decision part 20 decides a stage where code data and the code length in the buffer memory 15 are distributed based on an index (compared result) from the comparator 18, a reference value from a reference value calculation part 16 and a code length cumulative value from the buffer memory 15, and outputs it as a stage number.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing method and apparatus for encoding a frequency component of an image in block units and distributing the encoded data in a plurality of stages.

[0002]

2. Description of the Related Art In recent years, with the advent of color output devices, colorization of DTP for handling graphics, characters, images, etc. by a computer to create one image has been rapidly progressing. As a result, when a created color image is handled by a computer, etc., the amount of data is enormous. Therefore, a large amount of memory for storing the data and a large communication time cost for communicating it. Is required. In order to reduce the cost of the large-capacity memory and the communication time, a technique of compressing image data is indispensable, and various compression methods have been proposed.

As a general compression method of a color multi-valued image, there is an ADCT (Adaptive Discrete Cosine Transform) compression method by JPEG (Joint Photographic Expert Group), and the ADCT compression method is a compression method mainly for natural images. Although it is an irreversible compression method, a high compression rate can be expected. This ADCT compression method performs the following compression. That is, the three primary color (RGB) signals are converted into three components of Y, U and V. Here, Y represents a luminance component, and U and V represent chromaticity components. When the luminance component and the chromaticity component are compared, the luminance component is more sensitive to human vision than the chromaticity component, so that the Y component is compressed with the same resolution, and the U and V components are In some cases, subsampling reduces the resolution and compresses.

Next, Y obtained by subsampling
The UV data is subjected to DCT conversion for each block having a size of 8 × 8 for each component, and is extracted as a spatial frequency component. Hereinafter, the component converted by DCT will be referred to as "DCT coefficient". Then, these DCT coefficients are linearly quantized (divided) by a quantization table having a size of 8 × 8 set for each of the luminance component (Y) and the chromaticity component (U, V). Hereinafter, this quantized coefficient is referred to as a “quantized coefficient”. Finally, these quantized coefficients are coded using the Huffman coding method which is a variable length coding method.

The above is the compression procedure according to the ADCT compression method.

[0006]

However, since the Huffman coding method used in the above-mentioned ADCT compression method is a variable length coding method, the amount of compressed data is not known until the coding is completed, and therefore, There is a drawback in that it is not possible to control the target compressed data (compress the entire image data to a predetermined fixed length).

Further, this ADCT compression method is referred to as CG (compu
When applied to a character / graphic part created by ter graphics), there is a drawback that the error in subsampling or quantization becomes large and the deterioration of the character / graphic part becomes large. The present invention has been made to solve the above problems, and an object of the present invention is to provide an image processing method and apparatus capable of controlling the code data amount to a target data amount.

Another object of the present invention is to provide an image processing apparatus and method for monitoring the total code amount of sequentially input code data and allocating step information to the code data according to the total code amount. To do. A further object of the present invention is to perform a lossless encoding or an irreversible encoding adaptively in accordance with the characteristics of the input block image data and the characteristics of the image (for example, a character image and a halftone image), thereby making it appropriate. An object of the present invention is to propose an image processing apparatus and method capable of compressing an image to a code amount.

[0009]

The present invention for achieving the above object is an image processing apparatus for encoding the frequency coefficient of an image in block units and distributing the encoded data in a plurality of stages and outputting the encoded data. Encoding means for encoding the quantized frequency coefficient in block units, monitor means for monitoring the code amount of the code data output from the encoding means, and based on the code amount monitored by the monitor means, It is characterized in that it has an allocation means for allocating stage information to the code data from the encoding means, and an output means for outputting the code data from the encoding means according to this stage information.

According to the configuration of the invention relating to the image processing method of the present invention, the frequency coefficient of the image is coded in block units,
An image processing method for distributing the coded data in a plurality of stages and outputting the coded data, wherein a step of coding the quantized frequency coefficient in block units and a code amount of the coded data coded in the step are monitored. The method is characterized by including a monitoring step, a step of assigning stage information to the code data based on the code amount obtained by monitoring, and a step of outputting the code data according to the stage information.

According to another aspect of the image processing method of the present invention, an image is quantized in a frequency space, the quantized frequency coefficient is coded in block units, and the coded data is distributed to a plurality of stages and output. An image processing method, comprising an encoding step of encoding quantized frequency coefficients in block units, an addition step of adding code lengths of code data by the encoding step, and an addition result in the addition step. The present invention is characterized by including a step determination step of determining a step based on the step, and an output step of outputting the code data according to a determination result in the step determination step.

An image processing apparatus according to another aspect of the present invention is an image processing apparatus that compresses input image data that is input in block sequence by different compression methods. First image compression means for compressing the image data in a lossless compression method, and the block-sequential image data is compressed by the lossy compression method according to the compression result by the first compression means, and the compressed data is divided into a plurality of stages. A second compression means for distributing and outputting, a means for determining the feature of the block based on the compressed data obtained by compression by the first compression means, and a first feature based on the feature obtained by the determination means. And a means for selecting any one of the compressed data in the second compression means, and an output means for distributing the selected compressed data into a plurality of stages and outputting. .

According to another aspect of the image processing method of the present invention, an image is quantized in a frequency space, the quantized frequency coefficient is coded in block units, and the coded data is distributed to a plurality of stages and output. An image processing apparatus, wherein coding means for coding the quantized frequency coefficients in block units, counting means for counting coded data by the coding means, and steps based on the counting result by the counting means It is characterized in that it has a stage determination means for determining and the output means for outputting the code data according to the determination result by the stage determination means.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS A preferred embodiment of the present invention will be described in detail below with reference to the drawings. <Overall Configuration> FIG. 1 is a schematic block diagram showing the configuration of the image processing apparatus according to the present embodiment. In the figure, CPU11
Is for controlling the entire apparatus and setting various tables. The ROM 12 stores various tables and the like. The RAM 13 is a work area for setting a table and the like.

The case of compressing image data will be described below. First, the color conversion unit 1 converts RGB input image data into Y, U, and V components by 3 × 3 linear matrix conversion represented by the following equation (1). Here, Y represents a luminance component, and U and V represent chromaticity components.

[0016]

## EQU00001 ## Next, the sub-sampling unit 2 uses the fact that the sensitivity characteristic of the human eye is more sensitive to the luminance component (Y) than to the chromaticity component (U, V), and Subsampling is performed on the YUV signal from the unit 1. The sub-sampling compresses U and V with one of the following three compression ratios. The three ways are Y: U: V = 4: 4: 4
(No subsampling is performed), Y: U: V = 4:
Convert to 2: 2 (subsampling for U and V) or Y: U: V = 4: 1: 1 (subsampling for U and V). The output from the sub-sampling unit 2 is output in 8 × 8 block units, and the output order of the individual signals is Y: U: V = 4: 4: 4.
In the case of, Y1, U1, V1, Y2, U2, V2, and so on, and in the case of Y: U: V = 4: 2: 2, Y1,
Y2, U1, V1, Y3, Y4, U2, V2 ...
If Y: U: V = 4: 1: 1, Y1, Y
2, Y3, Y4, U1, V1, Y5, Y6, Y7, Y
It is output in the order of 8, U2, V2, ....

The DCT section 3 outputs these output data 8 ×.
DCT conversion is performed in units of 8 blocks and DCT coefficients are output. The quantizer 4 quantizes the DCT coefficient for each 8 × 8 block using the quantization table 8 and outputs the quantized coefficient. In addition, the quantizer 4 rearranges the two-dimensional quantized coefficients of 64 into one-dimensional ones in order from the low-frequency component to the high-frequency component according to the scan in the zigzag order as shown in FIG. It is sent to the conversion unit 5. The adaptive Huffman encoding unit 5 encodes the one-dimensional data in units of 64 by the method shown in FIG. 3 which will be described in detail later, and the encoded data and the length of the encoded data (expressed in the number of bits " Code length "), and a number (referred to as" stage number ") representing the distribution stage of the code data is output. still,
In the embodiment, the number of stages provided is set to 4 as an example.

The segment controller 6 writes the compressed data at the stage indicated by the "stage number" in the compression memory 7 based on the above-mentioned "code data", "code length" and "stage number". This compression memory is divided into segments (see FIG. 6), and the compressed data is written in the segment corresponding to each stage for each stage. Further, the information distributed to each segment is simultaneously written in the segment information table 10 (see FIG. 7). The data written in the information table 10 is used for decompression. Fixed-length compression can be realized by controlling the segment that stores the data sorted by stages by the segment controller 6.

The reason why this "fixed length compression" can be realized will be briefly described. As described above, the ADCT-transformed quantized coefficients are arranged in the order of DC component → low frequency AC component → high frequency AC component. "Lower stage number" is assigned to the code data of the lower frequency component,
When the "higher stage number" is assigned to the code data of the higher frequency component, outputting or storing more codes assigned the "higher stage number" means that the information of the original image is Allows you to save more accurately. That is, outputting or storing the code data of "higher stage number" means "compressing" with higher accuracy. Therefore,
The data distributed to each step is 1, depending on the size of the buffer memory and the amount of compressed data of the target image.
"Compress" in only two steps (output or store), or "compress" in only one, two, three steps
The target data amount can be selected by selecting the stage to be compressed (output or store) or “compress” (output or store) at all stages 1, 2, 3, and 4. Can be controlled. For example,
The amount of data that can be “output” in the first stage is 2.5 MB, the second
The amount of data that can be “output” in the stage is 1.5 MB, the amount of data that can be output in the third stage is 0.8 MB, the amount of data that can be “output” in the fourth stage is 0.5 MB, and the target amount of data is 5 If it is 0.0 MB, the total compressed data amount is 4.8 M by using the three data of the first, second and third stages.
B, the compression can be controlled within the target data amount. In this way, the accuracy of fixed-length compression is affected by the method of distributing all code data in stages. The above-mentioned segment controller 6, compression memory 7, and segment information table 10 will be described in detail later. <First Embodiment: Compressing Operation> Next, an outline of the adaptive Huffman encoding unit 5 will be described with reference to FIG. 3, and then details of each block will be described.

The outline of the Huffman coding unit 5 is as follows: "total target bit rate number" of the target of the entire image ("bit rate" is a unit indicating a compression rate and the number of bits per pixel [bits / pixel] ] And the target bit rate number of the first stage is set, after the encoding in the first stage is performed, the remaining target bit rate number in the block (= “total target bit rate number” − The "first stage target bit rate number") is roughly divided into three, and the equally divided bit rates are divided into the second, third, and fourth stages, respectively. , And "stage number" are output. For example, if the “total target bit rate” is 2.4 bits / pixel and the “first stage target bit rate number” is 1.5 bits / pixel for an 8 × 8 block image, the second, third, The “target bit rate number” in the fourth stage is 0.3, respectively.

Details of the adaptive Huffman coding unit 5 are shown in FIG. As shown in the figure, the adaptive Huffman coding unit 5 includes a Huffman coding unit 14, a buffer memory 15, a reference value calculation unit 16, a code length adder 17, a comparator 18, and a stage determination unit 20. . The encoding unit 5 outputs the code data, the code length, and the stage number as described above. The stage determination unit 20 determines which “stage number” is to be assigned to each piece of coded data encoded by the encoding unit 14. Code length adder 17, comparator 18, reference value calculation unit 16
Is the “index number” when the code length value of each code data is accumulated and the threshold value is tried to be exceeded based on the comparison between the accumulated value and a predetermined threshold value, and the “reference value” determined separately. Produces and. The determination unit 20 determines the “stage” based on the “index number” and the “reference value”. The buffer memory 15 synchronizes the “stage number” output from the determining unit 20 with the “code data” and the “code length” generated by the encoding unit 14 because it takes time to determine the stage. Required to output. The system of FIG. 3 will be further described below.

In FIG. 3, the Huffman coding unit 14
Huffman coding is performed for each of 64 quantized coefficients arranged one-dimensionally (for each Y, U, V component of 8 × 8 block), the coded data and the code length are stored in the buffer memory 15, and the block division is performed. The block end signal indicating is output to the code length adder 17. The buffer memory 15 stores these data in the form as shown in FIG.

The buffer memory 15 shown in FIG. 4 has an "index" field, a "code data" field for storing a Huffman code, a "code length" field for storing the length of the code, and a cumulative code length. It consists of a "code length cumulative value" field for storing the added value. The “index” indicates the number of the output code data in one block. In addition, the code length cumulative value is the code length adder 1
Obtained from 7.

In the example of FIG. 4, since the value of 0 to 18 is stored in the "index" field for the first block, it can be seen that the block has been encoded into 19 Huffman codes. Since the quantized coefficient input to the encoding unit 14 is DCT-transformed by the DCT unit 3, the first code data is a direct current (DC) component, and the subsequent code data is an alternating current (AC) component. Become. In the example of FIG. 4, the index 0 of the first block indicates the DC component, and the indexes 1 to 18 indicate the AC component.

The code length adder 17 cumulatively adds the code lengths of the code data of the AC component for each block in synchronization with the block end signal, and outputs the cumulative value to the comparator 18. The comparator 18 prepares threshold values to be compared with the accumulated value in the threshold table 19 for each of Y, U and V. FIG. 5 shows an example of the threshold table. In this example, each threshold is set to "40" for the Y component, "28" for the U component, and "28" for the V component. The comparator 18 outputs a reset signal to the adder 17. That is, the comparator 18 compares the cumulative value with the threshold value. When the cumulative value exceeds the threshold value, the index number of the previous code data (hereinafter referred to as "basic (Referred to as “index number”) is output to the stage determination unit 20 and the code length adder 17 is reset. The adder 17 starts adding the code length again from the code data next to the basic index number. When the adder 17 detects that the block has ended by the block end signal, the adder 17 calculates the reference value by using the cumulative value at that time (that is, the cumulative value of the code length until the block after reset is completed). Send to section 16.

The reference value calculator 16 calculates a plurality of reference values from the final cumulative value. This reference value is output to the determination unit 20, and the determination unit 20 compares this reference value with the comparator 1.
The step is determined based on the basic index number from 8. Hereinafter, the operation of the adaptive Huffman coding unit 5 in the first embodiment will be specifically described with reference to FIGS. 4 and 5.

FIG. 5 shows the case where the target bit rate in the first stage is set to 1.5 [bits / pixel],
This is an example of the threshold value for each of the Y, U, and V components, which are 40, 28, and 28 [bits], respectively. FIG. 4 shows code data, code lengths, and code length cumulative values for 8 × 8 blocks having Y components. As shown in the figure, since the index number 0 is a DC component, the adder 17 does not consider the code length of the DC component as an addition target, and the code length of the code data of the index = 1 (AC component) ( = 18) to write in the accumulated value field. Since the code length of the data of index number 2 is 7, the cumulative value is 18 + 7 = 25. Similarly, the code length cumulative value of index number 3 is 39.
Becomes Next, the code length of index number 4 is "21"
Therefore, the code length cumulative value is about to be "60". This "60" is compared with the threshold value (40 bits) of the Y component by the comparator 18, and 60> 40. The index number “3” of is sent to the stage determination unit 20 as the “basic index number”. Then, the code length adder 17 is reset by the comparator 18, and the code length "21" of the index number 4 is written as it is in the code length cumulative value. With index number = 5,
Since the code length is "9", the code length cumulative value is 21 + 9 = 30.

Similarly, the code length of each code data is added, and the result is written in the code length cumulative value field until the end of the block. Index = 18
In, the final value “95” of the code length cumulative value is transferred to the reference value calculation unit 16. The reference value calculation unit 16 divides 95 by 3 because the remaining number of stages is 3, and the quotient of “3
"0" and "60", which is twice that, are sent to the stage determination unit 20 as "determination reference values".

The stage determination unit 20 is based on the index number (the index number when the cumulative value is about to exceed the threshold value) obtained from the comparator 18 and the "judgment reference value" obtained from the reference value calculation unit 16. Thus, the “stage” to be given to each code data and code length data is determined. The operation of the determining unit 20 will be specifically described. As described above, the buffer memory 15 stores data as shown in FIG. From the buffer memory 15, “code data”, “code length” data and “code length cumulative value” are output in synchronization. "Code length cumulative value" of these three data
Is input to the determination unit 20. On the other hand, the decision unit 20
The “basic index number” from the comparator 18 is input. Therefore, the determination unit 20 gives a stage = 1 for the code data and the code length of the index before the “basic index number”. In the example of FIG. 4, since “basic index number” = 3, index numbers = 0 to 3
The code data and the code length data up to are treated as stage = 1. The deciding unit 20 decides and assigns a step number to the data of the index numbers after the "basic index number" based on the plurality of judgment period forward values. That is, the determination unit 20 sets “30” and “6” as the “determination reference value”.
0 ”is input from the reference value calculation unit 16. Therefore, the determination unit 20 compares the“ code length cumulative value ”read from the buffer memory with the“ judgment reference value ”(= 30) to obtain the“ code ”. "Code data" indicating that the "long cumulative value" is 30 or less
And "code length" data, a stage number = 2 is given, a stage number = 3 is given when 31 or more and 60 or less, and a stage number = 4 is outputted when 61 or more. .
Here, an EOS (End of Stage) code indicating the delimiter of the stage is always inserted between the code data at the delimiter of each stage.

As described above, it is possible to perform compression by dividing each block of 8 × 8 (= 64 coefficients) into a plurality of stages. In the above example, the bit rate of about 1.5 bits / pixel is assigned to the first stage data, and the bit rate is divided into three equal parts and compressed for the remaining data. In the examples of FIGS. 4 and 5 described above, the code length cumulative value of the Y component exceeds “40”,
When compressing image data of blocks that do not exceed the limit, an EOB (End of Block) code is inserted in the middle,
The 8 × 8 block code data are all written in the first-stage segment.

The "code data", the "code length" and the "stage number" thus obtained are sent to the segment controller 6 (FIG. 1), and the "code data" and the "code length" are transformed into the "stage number". Therefore, it is stored in the compression memory 7 in segment units. <First Embodiment: Decompression Operation> Next, the process of decompressing the data compressed by the above-mentioned process will be described below.

The system of FIG. 1 can operate in both compression and decompression. That is, the system of FIG. 1 operates as a decompressor only by reversing the flow of data when decompressing data. In this case, the DCT unit 3 is the inverse DCT unit 3 '.
, The quantizer 4 becomes the inverse quantizer 4 ', and
The adaptive Huffman coding unit 5 becomes an adaptive Huffman decoding unit 5 '. Also, the quantization table 8 and the Huffman table 9
Are the inverse quantization table and the Huffman decoding table, respectively, and the data flow is the reverse of that at the time of compression.

As described above, the code data forming one 8 × 8 block is divided into a plurality of stages and stored in the compression memory 7. Therefore, the adaptive Huffman decoding unit 5 ′ first requests the segment controller 6 for the code data of the stage number = 1. The segment controller 6 checks the contents of the segment information table 10, reads the first-stage code data from the compression memory 7, and transfers it to the adaptive Huffman decoding unit 5 '. The adaptive Huffman decoding unit 5'decodes the obtained code data one after another, transfers the result to the dequantization unit 4 ', and the EOS
The above process is repeated until (End of Stage) is detected.

When detecting the EOS, the decoding section 5'requests the controller 6 for the code data of the next stage number 2. As described above, the segment controller 6 checks the contents of the segment information table 10, reads the second-stage code data from the compression memory 7, and transfers it to the adaptive Huffman decoding unit 5 '. In the adaptive Huffman decoding unit 5 ', similarly to compression, EOB (End of B
lock) Decrypt until the code is detected. Similarly, the decoding of the third and fourth steps is performed, and the decoding of one 8 × 8 block is completed. In this case, EOB
, The next stage of code data is not requested,
Start decoding the next 8x8 block. If the compressed memory cannot be stored in the compressed memory until the final stage due to the amount of compressed data of the image and the capacity of the compressed memory, the data up to the intermediate stage is used for decompression.

The quantized coefficient thus obtained is inversely quantized by the inverse quantization section 4'using the inverse quantization table 8'and sent to the inverse DCT section 3 '. The inverse DCT unit 3'inverts the obtained DCT coefficient by inverse DCT to obtain Y'U '.
Obtain V'data. In the sub-sampling unit 2, the sub-sampling ratio (Y: U: V = 4: 4: 4, 4: 2:
2,4: 1: 1), the enlargement operation is executed. The color conversion unit 1 performs the inverse conversion according to the following equation (2) to restore the original image.

[0036]

<Equation 2><First embodiment ... Segmentation> Here,
Regarding the processing of storing the compressed data divided into multiple stages in the compression processing and decompression processing of the image data in the compression memory 7 composed of a plurality of segments and storing the segment information of the selected segment in the segment information table 10. The details will be described below.

The compression memory 7, as shown in FIG.
It is composed of a plurality of segments (for example, 1 segment = 100 kB) divided from 1 to SN. FIG. 7 shows the structure of the segment information table 10. The values of 2 to 11 ... In the segment information table 10 of FIG. 7 indicate the column numbers of the table, and indicate that the image goes from the beginning to the end as the image goes from left to right. Further, each of the four data in the first column of the table 10 indicates whether the corresponding stage is valid / invalid. In the example of FIG. 7, steps 1-3
Are valid, and stage 4 is invalid. The four data in each of the second to eleventh columns indicate the selected segment number (S-1 to SN) in the compression memory 7 in which the code data of each stage is written. Also, END
Indicates that the code data at each stage has ended.

Next, the compression data is compressed in the compression memory 7 at each stage.
The processing procedure for writing in, that is, to which segment the code data in each stage is assigned will be described. This allocation is :: The code data of the same stage for different pixels is
As long as there is free space, it will be stored in the same segment. : When the segment for storing the code data of a certain stage is full, the first segment that is not used by another stage is used. Such allocation makes it possible to use the compression memory 7 effectively without worming. This allocation will be specifically described.

FIG. 8 is a flow chart for explaining the compression data storage processing in the above compression processing. The compressed data distributed to each stage has a different code amount at each stage depending on the compression method for distributing the stages and the characteristics of the original image data. Here, when comparing stages 1 to 4, An example is used in which 1 outputs a larger code amount than 4 outputs.

Based on the above, the control procedure of the flowchart of FIG. 8 will be described below using the example of FIG. 7. In step S1 of FIG. 8, the code data output in each of the steps 1 to 4 for the image data of the first block is written in the segments S-1, S-2, S-3 and S-4, respectively. Thus, the process starts. The code data for each of the four stages occurring for the subsequent blocks are also segment S-1, S-2, S-3,
Will be written to S-4. This write operation continues in step S2 until it is detected that either segment is full. According to the compression method of this method, since the compression of the stage = 1 is always applied, the code data of the stage = 1 is generated in a larger amount than the other stages. Therefore, the segment S in which the code data of the stage = 1 is stored
-1 is expected to fill faster than segments S-2 through S-4. Therefore, if it is detected in step S2 that the segment S-1 is satisfied, step S3
Then, the writing of the code data of the stage = 1 starts in the first unused segment S-5.

Segments S-2 and S-3, which store code data for stages = 2 and 3, are filled when one of the empty areas is filled, because of which stage is full.
A new segment is assigned. In the example of FIG. 7, the segment S-2 for stage 2 is S-3 for stage 3.
Since it was full before, the segment for stage 2 was assigned to region S-6. Next, since S-3, which stores the data of the stage = 3, is full, the vacant S-7 is allocated.

Further, since the writing to the segment S-5 for writing the code data of the stage 1 is completed before the segment S-4 for storing the code data for the stage = 4 is full. , Free segment S
-8 is not assigned to stage = 4, but to stage = 1. This is because the code amount generated in stage 1 is
It happens because of more. Similarly, if the segment to be written is satisfied in each stage, the vacant segment is selected and the code data is written therein (steps S2 to S4).

In the embodiment, the method of coding the original image in several stages is used. Therefore, when the amount of encoded data at each stage is smaller than the target capacity of the compression memory 7, that is, when the compression of all images is completed before the segment SN is allocated. There is no memory capacity issue. However, when the last segment SN is allocated before the end of compression of all images (YES in step S4), more important code data,
That is, the segment used for the code data of the higher numbered stage (stage = 4 in the example of FIG. 7) is released for the coded data of the lower numbered stage (stage = 1 in the example of FIG. 7). Should be. Step S5 of FIG. 8 is for the release procedure.

That is, in the example of FIG. 7, the final segment S-
N was used to store the code data of stage = 2.
That is, at this time, there is no free area in the compression memory 7. Therefore, the segment storing the code data of the less important stage = 4 (S-4, S-1 in the example of FIG. 7).
1, S-15) are opened to use these segments as storage areas for code data for step = 1 or step = 2.
In the example of FIG. 7, segments S-4, S-11, S-15
Stores the code data of the stages = 1, = 2, and = 1 respectively. Then, the stage of the segment information table 10 =
Since only the invalid data is stored in the fourth row indicating 4, the corresponding bit in the first column is set to 0. In this way, when the compression memory 7 becomes insufficient, the less important step 4 is invalidated (for example, “0”), the segment S-4 used in step 4 is used for step = 1, Use segment S-11 for stage = 2. In the stage = 2, since the coding is completed in the segment S-11, the END mark is added thereafter. Further, in the stage = 1, the segment S-15 used in the stage = 4 is allocated, and since the encoding is completed thereafter, the END mark is added.

Regarding the decompression of the compressed data stored in the compression memory 7, the image data is decoded using only the valid stages 1 to 3 at the time of decompression. <Second Embodiment> Next, the configuration and operation of the adaptive Huffman coding unit 5 based on the index in the second embodiment will be described below with reference to the block diagram shown in FIG.

In the first embodiment described above, the step number is determined by accumulating the code lengths and determining whether or not the accumulated code lengths reach a predetermined value. In the second embodiment, the determination of stage = 1 is the same as the method of the first embodiment, but the determination after stage = 2 is performed based on the index number. In FIG. 9, the Huffman coding unit 21 sets the quantized coefficients arranged in one dimension to 64
Huffman coding is performed for each (for each Y, U, V component of 8 × 8 block), the code data and the code length are output to the buffer memory 22, and the code length data and the block end signal indicating the block delimiter are output. To the code length adder 25. The buffer memory 22 stores code data and code length data in a format as shown in FIG. In FIG.
The “index” is the same as the “index” number in the first embodiment, and indicates the number of the output code data in one block. The number of effective bits in the code data is the same as in the first embodiment. Specified by "code length".

The code length adder 25 adds and accumulates the code length of the code data (AC component only) for each block in synchronization with the block end signal, and outputs the accumulated value to the comparator 26. The comparator 26 compares a threshold value prepared for each of Y, U, and V with the cumulative value. This threshold value is substantially the same as the threshold value table 27 of the first embodiment.
Supplied from Comparator 2 when the added value exceeds the threshold value
When 6 determines, the index before the crossing is output to the counter 23 and the stage determination unit 24 as a "basic index". Thus, the "basic index" of the second embodiment has the same meaning as the "basic index" of the first embodiment.

When the threshold value table 27 has threshold values as shown in FIG. 5 and the code data is an example as shown in FIG. 10, the cumulative value is the threshold value (= 40) when the index = 4.
"Basic index" is "3"
Becomes In the second embodiment, the code data of the index before the “basic index” number, that is, the code data of the indexes 0 to 3 in the example of FIG. 10 is determined as stage = 1, and is stored in the compression memory 7 as stage = 1. Will be written.

Further, the counter 23 counts the number of all code data of one input block. That is, the counter 23 counts the index from the first AC component of one block. If the counter 23 detects that one block has been completed by the block end signal, a plurality of final index numbers (called “final index”) and the “basic index” obtained from the comparator 26 are used. The “reference indexes” are calculated and output to the stage determination unit 24.

In the example shown in FIG. 10, since the "basic index" number is "3" and the final index number is "18", the data to which the stages = 2, = 3, ... Are assigned is the index "." It can be determined from 4 "to" 18 ". Therefore, all code data of indexes 4 to 18 are divided into three equal parts, and stages = 2, = 3, ... Are assigned to each of the divided code data groups. Specifically, since (18-3) / 3 = 5, the number of codes in each group is "5", and the "reference index" is "8" (= 3 + 5) and "13" (= 8 + 5). ) Are output respectively. The block end signal from the Huffman coding unit 21 also resets the code length adder 25 and the comparator 26.

The stage determination unit 24 counts the index from the code length signal from the buffer memory 22. Thereby, the determination unit 24 can monitor the index number of the code data being output from the buffer memory. On the other hand, the determining unit 24 gives the “reference index” number (8, 13 in the example of FIG. 10) from the counter 23,
The "basic index" number (3 in the example of FIG. 10) is input from the comparator 26. In the example of FIG. 10, the determining unit 24 has a basic index of 3 and a reference index of 8,1.
3 is 3, the stage number = 1 for the data up to the index 3 shown in FIG. 10, the stage number = 2 for the data up to 4-8, and the stage number = 2 for the data up to 9-13. The stage number = 3 and the stage number = 4 are output until the end of this block. At that time, an EOS (end of stage) code is inserted at the transition of the stage number. This code is necessary for decompression.

The adaptive Huffman decoding unit for decompressing the code data by this coding method is the first
It can be realized by a method similar to that of the embodiment. <Third Embodiment> Next, the structure and operation of the index-based adaptive Huffman coding unit 5 in the third embodiment will be described below with reference to the block diagram shown in FIG. This third embodiment also monitors the number of input code data,
First, in the Huffman encoding unit 31, the quantized coefficients that are one-dimensionally arranged are set for every 64 pieces (each Y, U, 8 × 8 block).
Huffman coding is performed for each V component, and code data and code length are output.

Next, the counter 32 counts the code length signal. By this count, the counter 32 can know the number of input code data, that is, the index number. The counter 32 transfers the result, that is, the index, to the comparator 33. This comparator 33
Is compared with the value in the index table 35 as shown in FIG. The index table shown in FIG. 12 stores the number of code data to be assigned to each stage as an index number. That is, FIG.
2, the Y component is index number 1 to
3 to stage = 1, index numbers 4 to 8 to stage = 2, index numbers 9 to 15 to stage =
3, index numbers 15 and above are assigned to stage = 4. The comparator 33 sequentially compares the respective values from these index tables with the output of the counter 32. Then, each time the output value from the counter 32 exceeds the value in the table, a count-up signal is sent to the counter 34. The output of the counter 34 indicates the stage number, the initial value of which is 1, and when the count-up signal is input, the output value is incremented. That is, the stage number is increased by one.

When the data compressed by the compression method of the third embodiment is expanded by using the compression / expansion device shown in FIG. 1, the adaptive Huffman decoding unit 5'provides the segment controller 6 with the stage number 1 Request coded data. The segment controller 6 uses the segment information table 10
The code data of the first stage (number = 1) is read from the compression memory 7 by referring to the contents of the above and transferred to the adaptive Huffman decoding unit 5 ′. Adaptive Huffman decoding unit 5 '
Decodes the obtained coded data, refers to the index table 35, and outputs the next required stage number.
Similarly, the step number request and the decoding of the obtained code data are repeated until the number of decoded quantized coefficients becomes 64 (the number of elements of 8 × 8 blocks). When 64 pieces have been decoded, the next 8 × 8 block is decoded.

In the first embodiment (FIG. 3) and the second embodiment (FIG. 9) described above, the bit rate is classified according to the stage based on the code length. When the method of the example (FIG. 11) is used, the accuracy is lowered, but the algorithm is simple, the hardware scale is small, and it can be realized at low cost. Further, since it is not necessary to insert the EOS code at the stage delimiter of one 8 × 8 block, the amount of code data is reduced accordingly. <Modifications of First to Third Embodiments> In the above-mentioned embodiments, the number of steps for distributing compressed data is four, but
This can be any number, such as 2, 3, or 5, 6.

Further, the target bit rate of the entire image is set to 2.
Although it is set to 4 [bits / pixel], the number is not limited to this value and may be any number. Furthermore, the target bit rate of the first stage is 1.5 [bits
/ pixel], and the value of the threshold value table of FIG. 5 is determined based on it, but the value is not limited to this value, and any number may be used. Further, although the coded data after the second stage is distributed so as to be divided into about three equal parts, the present invention is not limited to this, and the ratio of the second, third and fourth stages is, for example, 3: 2: 1, 5: 3: 1, etc. You can divide it in any way. In addition, in the third embodiment, as an index table serving as a reference for the step allocation, FIG.
However, the value is not limited to this. <Fourth Embodiment> Next, a fourth embodiment of the present invention will be described with reference to the drawings.
The embodiment will be described in detail.

In the fourth embodiment, the characters and figures included in the color image are compressed by the reversible compression method, and the natural image is compressed by the lossy compression method in the first to third embodiments described above. The compressed data is sorted into a plurality of stages and stored. Figure 13
It is a schematic block diagram which shows the structure of the image compression apparatus concerning 4th Example. In the figure, reference numeral 41 is a raster-to-block converter, which is 2 per pixel that is input in raster order.
4-bit pixel data is output in block (8 × 8 pixels) units. A block delay unit 42 delays the image data by temporarily storing the image data for one block. A lossy compression unit 43 compresses image data by a lossy compression method, compresses the block sequential data from the block delay unit 42, divides the compressed code data into a plurality of stages, and outputs the divided data. Reference numeral 44 denotes a segment controller, which can store the code data divided into multiple stages in the first storage unit 45 divided into a plurality of segments, as in the above-described embodiment. By this storage means, fixed length compression can be realized, and the segment number selected for each stage is stored in the segment information table 46.

The lossless compression unit 47 compresses the image data input in block order in a block unit in a lossless manner, stores the result in the compressed data holding unit 48, and counts the amount of the compressed data. Then, the count result is D
When the limit value allowed for the block is L, and L ≧ D, the block is set to area 0, and L
If <D, the block is set to area 1. Then, the result (“determination signal”) is stored in the second storage unit 49.
Therefore, the second storage unit 49 is a block map in which one block is represented by 1 bit that distinguishes the area 0 or the area 1. That is, the block determined to be the area 1 is determined to be the block whose code amount D exceeds the limit value L.

In the compression system of the fourth embodiment, for blocks (area 0) in which the compression code amount compressed by the lossless encoding method does not exceed the limit value L, the code by the lossless encoding method is used. The block (area 1) that has been adopted and exceeds L is lossy-encoded. In FIG. 13, the lossy compression unit 43 encodes the data delayed by one block, and therefore, the input of the segment controller 44 is the lossless-compressed code data of the image data of one block and the lossy-compressed code. Synchronized with data. In addition, the lossless compression unit 47
Sends the above-mentioned determination signal to the compressed data holding unit 48 and the lossy compression unit 43. That is, the lossless compression unit 47 causes the segment controller 44 to transfer the data of the compressed data holding unit 48 in which the compressed data is stored for the block determined to be the area 0. Further, regarding the block determined as the area 1, the compressed signal stored in the compressed data holding unit 48 of the block is invalidated, and the instruction signal is lossy compressed in order to compress the block by the lossy compression method. The data is transferred to the unit 43 and compressed by the lossy compression unit 43. The lossy compression unit 43
The configuration and the operation of are the same as those of the above-described embodiment, and the description thereof is omitted here.

Here, color DTP is added to the system of FIG.
Consider a case where a color image including characters, figures, natural images, etc. created by is input. Since CG images such as characters and figures have high compression efficiency when compressed by the reversible compression method, D <L is satisfied, and it is preferable that the CG image is compressed by the reversible compression method which is excellent in restoration property. On the other hand, since a natural image has a large variation in pixel value, even if lossless compression is performed, the data amount cannot be expected to be small. Therefore, it is desirable to compress a natural image using a lossy compression method with high compression efficiency.

According to the process of the fourth embodiment, the color D
A color image including characters, figures, natural images, etc. created by TP can be compressed by a lossless compression method for CG images of characters, figures, etc., and a lossy compression method for natural images. Next, the process of decompressing the compressed data will be described below with reference to FIG.

FIG. 14 is a schematic block diagram showing the arrangement of an image decompression device according to the fourth embodiment. In the figure, FIG.
The same symbols are attached to the same components. This decompression device performs decompression based on the data stored in the first storage unit 45 and the second storage unit 49 by the system of FIG.
The segment controller 44 generates an address of the first storage unit 45 divided into a plurality of segments from the segment information table 46 according to the code data and the code length requested by the lossy decompression unit 51 or the lossless decompression unit 52, The code data is transferred to the lossy decompression unit 51 or the lossless decompression unit 52.

The irreversible decompression unit 51 is for decompressing the data compressed by the irreversible compression unit 43 of FIG. 13, and generates the code length and the stage number of the code data requested by itself. Then, the obtained coded data is decompressed and transferred to the switch unit 53 in block units. Reversible extension 52
Is for decompressing the data compressed by the lossless compression unit 47 of FIG. 13, and generates the code length and the stage number of the code data requested by itself. Then, the obtained coded data is decompressed and transferred to the switch unit 53 in block units.

The switch section 53 selects and outputs input data based on the bit information stored in the second storage section 49. For example, when the bit information is “0”, the data from the lossless decompression unit 52 is selected, and when the bit information is “1”, the data from the lossy decompression unit 51 is selected. Then, the block raster conversion unit 54 converts the block-sequential data output from the switch unit 53 into raster-sequential data, and the decompression process ends.

According to the fourth embodiment, the lossless compression method and the irreversible compression method are used together to prevent the deterioration of the character / graphic part created by CG, and the lossless compression method allows the compressed data to be compressed according to the limit value. In the lossy compression method, the amount of compressed data can be controlled by dividing output compressed data in multiple stages and storing it in a memory divided into a plurality of segments. <Modification of Fourth Embodiment> The reversible compression section 47 described above compares the means for counting the compressed data according to the reversible compression method, the means for setting the limit value L, and the count result D and the limit value L for comparison. Various modifications are possible as long as it is provided with means for outputting the result.

In the fourth embodiment, when the compressed data divided into multiple stages is stored in the compression memory 7 divided into a plurality of segments, one having a large number of stages (stage 4) to one having a small number of stages (stage 1) However, the present invention is not limited to this and may be adaptively selected and invalidated. Further, although the type shown in FIG. 11 is used as the segment information table, the segment information table is not limited to this and various modifications can be made without departing from the scope of the present invention.

The present invention may be applied to a system composed of a plurality of devices or an apparatus composed of a single device. Further, although all the above-described embodiments have their operations realized by hardware logic, it goes without saying that they can also be applied when they are achieved by a program.

There is also no.

As described above, since the separating means is not required in the temporary storage means, it is possible to reduce the size of the entire system and reduce the cost.

[Brief description of drawings]

FIG. 1 is a schematic block diagram showing the configuration of an image processing apparatus in this embodiment.

FIG. 2 is a diagram showing a scanning direction in a zigzag scanning operation.

FIG. 3 is a block diagram showing a configuration of an adaptive Huffman coding unit in the first embodiment.

FIG. 4 is a diagram showing data stored in a buffer memory 15 in the first embodiment.

FIG. 5 is a diagram showing a threshold table.

FIG. 6 is a diagram showing a configuration of a compression memory divided into segments.

FIG. 7 is a diagram showing a structure of a segment information table.

FIG. 8 is a flowchart showing a storage process of compressed data.

FIG. 9 is a block diagram showing a configuration of an adaptive Huffman coding unit in the second embodiment.

FIG. 10 is a diagram showing data stored in a buffer memory 22 in the second embodiment.

FIG. 11 is a block diagram showing a configuration of an adaptive Huffman coding unit in the third embodiment.

FIG. 12 is a diagram showing an index table.

FIG. 13 is a block diagram showing a configuration of an image compression apparatus in a fourth embodiment.

FIG. 14 is a block diagram showing a configuration of an image decompression device according to a fourth embodiment.

[Description of Codes] 1 color conversion unit 2 sub-sampling unit 3 DCT unit 4 quantization unit 5 adaptive Huffman encoding unit 6 segment controller 7 compression memory 8 quantization table 9 Huffman table 10 segment information table 11 CPU 12 ROM 13 RAM 14 Huffman coding unit 15 Buffer memory 16 Reference value calculation unit 17 Code length adder 18 Comparator 19 Threshold table 20 Step determination unit 21 Huffman coding unit 22 Buffer memory 23 Counter 24 Step determination unit 25 Code length adder 26 Comparison 27 Threshold table 31 Huffman coding unit 32 Counter 33 Comparator 34 Counter 35 Index table

Claims (24)

[Claims] 1. An image processing apparatus for encoding a frequency coefficient of an image in block units and distributing the encoded data in a plurality of stages and outputting the encoded data, wherein the quantized frequency coefficient is encoded in block units. Means, monitor means for monitoring the code amount of the code data output from the encoding means, and based on the code amount monitored by the monitor means,
An image processing apparatus comprising: an assigning unit that assigns stage information to the code data from the encoding unit; and an output unit that outputs the code data from the encoding unit according to the stage information. . 2. The monitor means stores, for each stage, the number of codes as a threshold value assigned in advance to the stage, and the monitor monitors the number of code data output from the encoding means. It is provided with means for comparing the number of the code data inputted from the means with the threshold value read from the storage means, and means for determining the stage information according to the comparison result by the comparing means. Item 1. The image processing device according to item 1. 3. The monitor means stores, for each stage, a code amount as a threshold value previously assigned to the stage, the total code amount from the monitor unit, and the storage means. The image processing apparatus according to claim 1, further comprising: a unit that compares the read threshold value with a unit that determines the stage information according to a comparison result by the comparing unit. 4. The data storage means, which is divided into segments of a predetermined unit size in advance for storing a plurality of code data, and the free capacity of the segment for each segment. 2. The image processing according to claim 1, further comprising: a storage unit, a table unit that stores the correspondence between each segment of the data storage unit and the stage to which the code data stored in each segment belongs. apparatus. 5. The output means in one segment,
When the monitoring means determines that there is no more area for newly storing code data belonging to the stage assigned to the segment, an unused segment in the data storage means is searched for, and the new code is added to the segment. The image processing apparatus according to claim 4, further comprising: a unit for storing data and updating the data of the table unit. 6. The image processing apparatus according to claim 4, wherein the output unit, when searching for an unused segment, assigns the segment closest to the segment in use. 7. The output means releases the segment storing code data belonging to a later stage when the empty area of the data storage means is exhausted. The image processing device described. 8. The output means includes one code data,
The image processing apparatus according to claim 1, wherein the coded data is output as a set with the stage information assigned by the assigning unit. 9. The image processing apparatus according to claim 1, wherein a frequency component of the image is extracted by orthogonal transformation, and the frequency component is encoded by Hufmann code. 10. The image processing apparatus according to claim 1, wherein the frequency component of the image is input in a known frequency sequence and encoded by the encoding means. 11. The image processing apparatus according to claim 10, wherein the frequency components of the image are arranged in order from a low frequency component to a high frequency component. 12. An image processing method for encoding a frequency coefficient of an image in block units and distributing the encoded data in a plurality of stages and outputting the encoded data, wherein the quantized frequency coefficient is encoded in block units. , A monitoring step of monitoring the code amount of the code data encoded in the above step, a step of assigning stage information to the code data based on the code amount obtained by monitoring, and a code according to the stage information. And a step of outputting data. 13. The monitoring step stores, for each stage, the number of codes as a threshold value previously assigned to the stage, and inputs the number of inputted codes as the total amount of codes, The image processing method according to claim 12, wherein the number of codes is compared with a read threshold value, and the stage information is determined according to the comparison result. 14. The monitoring step stores, for each stage, a code amount as a threshold value previously assigned to the stage, compares the total code amount with the read threshold value, and compares the total amount. 13. The image processing method according to claim 12, wherein the stage information is determined according to the result. 15. The output step includes the step of monitoring the free space of each segment of the data storage means that has been divided into segments of a predetermined unit in advance, and the step of assigning one segment to that segment. When it is determined that there is no more area for newly storing the code data belonging to, the unused segment of the data storage means is searched, the new code data is stored in the segment, and the data storage means is stored. 13. The image processing method according to claim 12, further comprising the step of updating the data of the table storing the correspondence between each segment and the stage to which the coded data stored in each segment belongs. 16. The image processing method according to claim 15, wherein in the output step, when an unused segment is searched for, a segment closest to the segment in use is assigned. 17. The output process according to claim 15, wherein the segment storing the code data belonging to a later stage is released when the empty area of the data storage means is exhausted. The described image processing method. 18. The image processing method according to claim 12, wherein in the output step, one piece of code data and stage information assigned to the code data are combined and output. 19. The image processing method according to claim 12, wherein a frequency component of the image is extracted by orthogonal transformation, and the frequency component is encoded by Hufmann code. 20. An image processing method for quantizing an image in a frequency space, coding the quantized frequency coefficient in block units, distributing the coded data in a plurality of stages, and outputting the coded data. An encoding step of encoding coefficients in block units; an addition step of adding the code lengths of code data by the encoding step; a step determination step of determining a step based on the addition result in the addition step; And an output step of outputting the coded data according to the determination result of the step determination step. 21. An image processing apparatus for compressing input image data, which is input in block sequence, by different compression methods, wherein the input image data in block sequence is compressed by a reversible compression method. Compressing the block-sequential image data by a lossy compression method according to the compression result by the first compressing means, and distributing the compressed data in a plurality of stages and outputting the compressed data. Which of the means for judging the feature of the block based on the compressed data obtained by the compression by the first compressing means and the compressed data for the first and second compressing means based on the characteristic obtained by the judging means An image processing apparatus comprising: means for selecting whether or not and output means for distributing the selected compressed data in a plurality of stages and outputting the data. 22. The image processing apparatus according to claim 21, wherein the determination unit determines a character, a graphic, or a natural image. 23. A character, a graphic or a natural image is compressed and output by the first compression means, and a natural image is compressed by the second compression means, and the compressed data is compressed. 23. The image processing apparatus according to claim 22, wherein the image processing apparatus distributes and outputs to a plurality of stages. 24. An image processing apparatus which quantizes an image in a frequency space, encodes the quantized frequency coefficient in block units, distributes the coded data in a plurality of stages, and outputs the quantized frequency. Coding means for coding coefficients in block units, counting means for counting code data by the coding means, stage judging means for judging a step based on the counting result by the counting means, and the step judging means. And an output unit that outputs the coded data in accordance with the determination result in 1.