JPH07283948A - Compressing device for gradation picture data - Google Patents

Abstract

PURPOSE:To obtain a restored picture of high picture quality with a high compression rate by using frequencies in occurrence of quantized conversion coefficient AC components as a parameter to classify blocks into several classes and using the quantization width adapted to each class to compress the class. CONSTITUTION:A reader 2 reads out picture data from a frame memory 1 in block units. With respect to read-out block data, a conversion coefficient of a one-block portion transformed by a two-dimensional discrete cosine transforming device 3 is obtained. Conversion coefficient AC components except one DC component out of conversion coefficients obtained in this manner are uniformly quantized with a fundamental quantization width by a quantizer 4, and a pair of a coefficient number as a fixed length code and a polarity number are obtained for each conversion coefficient. A classifying device 5 performs classification processing in accordance with the coefficient number of the one-block portion sent from the quantizer 4 and the compression rate parameter inputted from a terminal 52. An encoder 7 converts the fixed length code of the conversion coefficient AC component to a variable length code and outputs it.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a gradation image data compression apparatus for compressing gradation image data such as medical images such as X-ray photographs and CT images.

[0002]

BACKGROUND OF THE INVENTION With the progress of digital technology, it has become more frequent to digitize and store and transmit gradation images and to perform various digital image processing. However, the gradation image has a large amount of information as compared with the binary image, and therefore the amount of data when digitizing the gradation image is a problem. Especially in medical images,
The number of pixels for digitization and the number of bits required for each pixel are, for example, 4 million pixels for chest radiographs, 8 to 8
It is as large as 10 bits, which is inefficient in saving and transferring data.

Therefore, today, a data compression technique for compressing enormous amount of data of a gradation image including a medical image to make it compact is in the limelight.

Data compression techniques are roughly classified into lossless compression and lossy compression.
Since only a compression rate as low as / 3 can be expected, an irreversible compression method that can obtain a compression rate as high as ⅕ or more, in particular, a conversion coding method is drawing attention.

Transform coding is one of the lossy compression schemes in which the entire image is divided into small blocks, orthogonal transformation is performed in block units, and the transform coefficients obtained thereby are quantized and coded. This is the most suitable compression method for compressing toned images.

In transform coding, it is known that the distribution of the AC component of the transform coefficient is close to a Gaussian distribution having a peak at zero, and the quantization of the transform coefficient having such a distribution is quantized. Mid-Ris with judgment level including zero
er-type quantization and Mid including zero in the quantized output level
-Trace type quantization, Mid-tr
It has been reported that the ace type quantization is more preferable because it has less random noise in the block. However, when the Mid-trace type quantizer is used, there is a drawback that a pattern image peculiar to the component appears when the compression rate is increased, which is unnatural when image processing such as enlargement, gradation, and frequency processing is performed. There is a problem in that the image appears and the fidelity of the restored image is low.

By the way, it is understood that the image quality of the restored image in the case of performing the mid-trace type quantization has a strong correlation with the number of transform coefficient AC components quantized to zero in the quantization step during compression. That is, since the image of each block of the restored image is the weighted sum of the pattern peculiar to the component of the conversion coefficient AC component quantized to other than zero at the time of quantization, the conversion coefficient AC component quantized to other than zero When the number of s is small, in other words, when the number quantized to zero is large, a component-specific pattern image appears in the restored image. Therefore, in order to suppress the appearance of the pattern image peculiar to the component, the number of transform coefficient AC components quantized to zero is reduced by narrowing the quantization width of the quantization determination level at which the quantized output level becomes zero. In other words, it is sufficient to increase the number of quantized values other than zero, but if this is done, the average code length at the time of encoding increases and the compression rate decreases.

Therefore, the occurrence frequency of the transform coefficient AC component quantized to zero is examined for each block. Since the number of all conversion coefficient AC components in the block is fixed,
It is similar to checking the number other than zero. By doing so, it was found that a block in which the pattern image peculiar to the component is likely to appear is detected, and the quantization width is narrowed only in the detected block to obtain a restored image in which the appearance of the pattern image is suppressed.

[0009]

SUMMARY OF THE INVENTION The present invention has been made in view of the above points, and in compressing data of a gradation image including a medical image by using a transform coding method for performing Mid-trace type quantization. , By using a classification method for preventing the occurrence of a pattern image peculiar to a component and optimally classifying blocks, in order to obtain a high-quality restored image with a high compression rate, in order to achieve this purpose, Blocks are classified into several classes with the frequency of occurrence of the transform coefficient AC component that is quantized to zero in the blocks when Mid-trace type quantization is performed as a parameter, and the quantization width suitable for each class Is used for compression.

[0010]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing an embodiment of a gradation image data compression apparatus according to the present invention.

In the figure, a frame memory 1 stores gradation image data to be compressed (in this example, the number of bits per pixel is 8 bits). The image data is read in blocks from (in this example, the block size is 16 pixels in the line direction and 16 pixels in the column direction). From the read block data, 256 transform coefficients for one block, which are cosine transformed by the two-dimensional discrete cosine transform (2D-DCT) device 3, are obtained.

Next, of the 256 transform coefficients thus obtained, the 255 transform coefficient AC components excluding one DC component are uniformly quantized by the quantizer 4 with the basic quantization width w _0. , A pair of a coefficient number and a polarity number which is a fixed length code is obtained for each transform coefficient. The pair of coefficient number and polarity number is temporarily stored in the block buffer memory 6 and the coefficient number is sent to the classifying device 5.

The classifying device 5 performs a classifying process on the basis of the coefficient number for one block sent from the quantizing device 4 and the compression ratio parameter input from the terminal 52, and assigns the class number to the encoding device 7. Output. When the coding device 7 reads the class number sent from the classifying device 5, it reads a pair of coefficient number and polarity number for one block from the block buffer memory 6, and a fixed length code (number) of the transform coefficient AC component. Is converted into a variable length code and output to the terminal 8. The code data is transmitted from the terminal 8 or stored in the memory.

FIG. 2 shows an example of the quantizer 4.

First, from the terminal 41, 255 conversion coefficient AC components excluding the DC component of the conversion coefficient for one block are input. The absolute values of the 255 conversion coefficients are taken by the absolute value circuit 42, and at the same time, the polarity judgment circuit 43 judges whether the conversion coefficients are positive or negative. Polarity determination circuit 43
Outputs the determination result to the terminal 47 as a polarity number. Here, the polarity number is represented by a 1-bit code, and is "1" when the conversion coefficient is negative and "0" when the conversion coefficient is positive.

On the other hand, the absolute value of the transform coefficient output from the absolute value circuit 42 is divided by the basic quantization width w ₀ by the division circuit 44. The division result is truncated by the truncation circuit 45 after the decimal point (this result is called a coefficient number), and is output to the terminals 46 and 47. The terminal 46 is connected to the classification device 5, and the terminal 47 is connected to the block buffer memory 6. Terminal 41
The polarity number and the coefficient number obtained for one transform coefficient input from the pair form a pair and are temporarily stored in the block buffer memory 6 from the terminal 47.

On the other hand, the DC component of each block is either uniformly quantized with a sufficiently small quantization width w _dc , or is quantized without being quantized, and like the other 255 transform coefficient AC components, the block buffer memory 6 is used. Is temporarily stored in.

FIG. 3 shows the quantization of the conversion coefficient AC component excluding the DC component.

The horizontal axis represents the value of the conversion coefficient, and the vertical axis represents the frequency of occurrence of the conversion coefficient. The number sequence shown in FIG. 3A is the coefficient number (k) sequence output from the truncation circuit 45, and the number sequence shown in FIG. 3B is the polarity number (j) output from the polarity determination circuit 43. ) Column.

Next, FIG. 4 shows an example of the classifying device 5 when the number of classes is four.

The terminal 51 is connected to the terminal 46 in FIG. 2, and the coefficient number image output from the quantizer 4 is input from here. Here, the coefficient number is k (k = 0, 1, 2,
3, 4, ...), the coefficient number k input from the terminal 51 is converted into a 4-bit primary counter control code C _i (i = 0, 1, 2, 3) by the counter control LUT 53.

An example of controlling the correspondence between each block and its quantization width by the parameter input of the present invention will be described using the example of the classification device of FIG.

FIG. 5A shows an example of the counter control LUT 53.

On the other hand, the compression ratio parameter P input by the user from a terminal (not shown) is input from the terminal 52. That is, the parameter P is given to the classifying means. In this example, two compression ratios are prepared (P
= 0, 1), which means that the larger P is, the higher the compression rate is. The compression ratio parameter P input from the terminal 52
Is converted into a 4-bit compression rate control code e _p by the compression rate control LUT 54.

FIG. 5B shows an example of the compression rate control LUT 54.

Next, the OR circuit 55 fetches the primary counter control code c _i and the compression ratio control code e _p, and the 4-bit secondary counter control code d _i, which is the logical sum of the two, as in equation (1 _). Ask for.

[0028]

[Number 1] d _i ← c _i ∪ e _p then the counter control circuit 56 takes in the secondary counter control code d _i, the four counters from the counter 0 in the counter circuit 57 to the counter 3 of d _i O for each bit
Controlled by n / off state. Each bit of the secondary counter control code d _i is associated with the counter 0, the counter 1, the counter 2, and the counter 3 in order from the lower bit, and when the bit is on (= 1), the counter control circuit 56 corresponds. Increment the counter to
When the bit is off (= 0), the corresponding counter is not counted up.

FIG. 5C shows the correspondence between each bit of d _i and each counter. Note that the same correspondence relationship holds for each counter for c _i and e _p .

FIG. 6 is a diagram showing the relationship between the compression ratio parameter P or the compression ratio control code e _p and the operating range of each counter.

For each compression ratio parameter P (each e _p ), each counter operates only for the coefficient number k in the range indicated by the arrow in FIG. However, in this example, the counter 3 operates for all coefficient numbers k regardless of the value of P, so that the counter 3 can know the end of the classification work for one block. Therefore, when the other counter similar to the counter 3 is owned, the counter 3 may be omitted.

Next, the comparison processing procedure of the comparison circuit 58 is shown in FIG.
It is shown in the flowchart.

The comparison circuit 58 constantly checks the count value of the counter 3 (F-1), and the count value is 255.
If less than, the comparison circuit 58 is in a standby state. When the count value of the counter 3 reaches 255, the comparison circuit 58 outputs the count threshold value S output from the count threshold value LUT 59.
_{Read p} .

FIG. 5D shows an example of the count threshold LUT 59.

The count threshold value S _p is a threshold value for the count value issued for the compression rate parameter P input from the terminal 52. When the comparison circuit 58 reads the count threshold value S _p , this S _p is compared with the count value of the counter 2 (F-2). As a result, (1) If the count threshold value S _p is greater than or equal to the count value of the counter 2, the class number m is set to 3 (F-
3) If S _p is smaller than the count value of the counter 2, S _p
p is compared with the count value of the counter 1 (F-4). (2) In the comparison between the count threshold value S _p and the count value of the counter 1, if S _p is equal to or larger than the count value of the counter 1, the class number m is set to 2 (F-5), and S _p is the counter value of the counter 1. If it is smaller than the count value, S _p is compared with the count value of the counter 0 (F-6). (3) In the comparison between the count threshold value S _p and the count value of the counter 0, if S _p is greater than or equal to the count value, the class number m is set to 1 (F-7), and S _p is greater than the count value of the counter 0. If it is smaller, the class number m is set to 0 (F-8). (4) When the class number m is set, the class number m is output from the terminal 510 (F-9), the classification work for one block is completed, and the counters 0 to 3 are counted. Clear the value to zero (F-1
0). The class number m output from the comparison circuit 58 is a parameter that indicates the range of transform coefficients that are run-length encoded at the time of encoding. That is, the class number is m
It means that the transform coefficients existing in the region where the absolute value of the transform coefficient is 0 or more and _Am or less are run-length encoded with respect to the 255 AC components of the block (see FIG. 6).

FIG. 8 shows an example of the encoding device 7.

When the classifying device 5 finishes the classifying work for one block, the coding device 7 reads the class number m from the terminal 79 (the terminal 79 is connected to the terminal 510 in FIG. 4) and quantizes it. One of the four quantization tables stored in the LUT 73 is used for quantization.
One of them is selected, and at the same time, a data read command from the block buffer memory 6 is sent to the data read circuit 72. The data read circuit 72 reads the DC component data and 255 pairs of coefficient number k and polarity number j (j = 0 or 1) from the terminal 71 connected to the block buffer memory 6. If the read data is a DC component, the data is output from the terminal 78, and if the read data is a pair of the coefficient number k and the polarity number j, the quantization in the quantization LUT 73 selected by the class number m. The coefficient number k and the polarity number j are marked by the table.

FIG. 9 shows an example of the quantization table in the quantization LUT 73.

This example is an example in which the coefficient number k is evenly divided so that the final quantization state becomes the uniform quantization shown in FIG. 10, but in addition to this, the coefficient number k is unevenly divided. It is also possible to perform non-uniform quantization with a final quantization width that is an integral multiple of the basic quantization width w ₀ .

Next, an example of the encoding procedure of the AC component is shown in FIG.
It is shown in the flowchart.

[0041] Is sent to the zero decision circuit 74 having the data 74a (P-
1). Code passage counter After counting up (P-4), whether the code passage count value y becomes 255 Is output (P-7), the code passage count value y is 255.
It is determined whether or not (P-11). If the zero count value z is not zero, sign zero count value z
Converted to variable length code r _z by UT77 (P-8),
After zero-clearing the zero count value z (P-9), r _z
Is output from the terminal 78 (P-10), and it is determined whether the code passage count value y reaches 255 (P-1).
1).

In the judgment of the code passing count value y, if the count value y is not 255 To do. If the code passage count value y is 255 (P-11), it is determined whether the zero count value z is zero (P-12). If it is zero, the code passage count value y is cleared to zero (P-12). P-13) The coding operation of the transform coefficient for one block is completed. If the zero count value z is not zero, a code conversion step (P-8), a zero clear step of the zero count value z (P-9), and a terminal 78
After passing through the variable length code r _z output step (P-10) to 0, the code passage count value y is cleared to zero (P-13), and the coding operation of the transform coefficient for one block is completed.

FIG. 12A shows an example of the code LUT 76.

[0044] This enables a simpler device.

FIG. 12B shows an example of the code LUT 77.

The present example is an example in which the run length code assigned to the zero count value z uses the B1 code common to each class. When the gradation image data is compressed by this method, the frequency distribution of each zero run length is an exponential distribution suitable for the B1 code, so that it is possible to obtain a high compression rate by using the B1 code.

In FIGS. 12A and 12B, C is a 1-bit code having a value of 0 or 1.

FIG. 13 shows the data read order of the data read circuit 72 in FIG.

Since the conversion coefficient of the gradation image tends to have a smaller amplitude in the higher frequency component, the probability of occurrence of a component quantized to zero becomes higher in the higher frequency component. Therefore, as shown in FIG. 13A or 13B, if coding is performed from the low frequency side toward the high frequency direction, the probability that a long zero run will occur is increased, and effective run length coding can be performed.

FIG. 14 shows an example of the code structure.

FIG. 14A shows a code example as a result of compressing the entire image of one frame, which is composed of a header part, a code of each block following the header part, and a data end code.

FIG. 14B shows an example of the header structure, which stores the block size (16 in this example), the number of blocks in the image line direction, the number of blocks in the image column direction, and the compression ratio parameter P.

FIG. 14C shows an example of the code structure of each block. The class number m (2-bit fixed length code in this example), the DC component code (fixed length code), and the following AC component code string ( Variable length code string). The AC component code string is as shown in FIG. r _z ) are arranged alternately.

An example of various parameters used in this embodiment is shown below.

Basic quantization width w ₀ = 0.5 A ₀ = 0.5, A ₁ = 1.0 A ₂ = 1.5, A ₃ = 2.0 S ₀ = 200, S ₁ = 225 The parameter is the amount of information per pixel of the original image is n
The block size is N × N in bits (n = 8 in this example)
When (N = 16 in this example), the block image is f (x,
y) is a value when the two-dimensional discrete cosine transform formula (shown in Formula 2) in which the transform coefficient matrix is represented by F (u, v) is calculated at the floating point.

[0056]

[Equation 2] In the present invention, instead of the two-dimensional discrete cosine transform device 3, another orthogonal transform device such as a two-dimensional discrete Hadamard transform device or slant transform device may be used. Further, in the present invention, the block size is not limited to the size of 16 × 16, and the size of 8 × 8, 32 × 32, 64 × 64, etc. can be applied, and the number of bits per pixel of the original image is also applicable. It is not limited to 8 bits. Further, in the classifying device 5, when counting the coefficient numbers k in a specific section, the classification may be performed based on the number of coefficient numbers k existing in one specific section without providing a large number of specific sections. It is possible. It is also possible to combine the classification parameters of the present invention with other classification parameters for classification.

[0057]

As described above, in the present invention, by using the occurrence frequency (the number) of the conversion coefficient AC component quantized to zero as one of the classification parameters, the image quality of the restored image, In particular, classification based on the degree of occurrence of component-specific pattern images, which is a drawback of Mid-trace type quantization, becomes possible. For example, by determining the quantization width so that the frequency of a block with high frequency is low, the pattern image It is possible to suppress the occurrence of the above and obtain a high-quality restored image.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of a gradation image data compression device according to the present invention.

FIG. 2 is a block diagram of an example of the quantization device shown in FIG.

FIG. 3 is a diagram showing a relationship between a coefficient number and a polarity number obtained by quantizing a conversion coefficient AC component in relation to a conversion coefficient occurrence frequency.

FIG. 4 is a block diagram of an example of the classification device shown in FIG.

5A is a diagram showing a counter control LUT, FIG. 5B is a compression ratio control LUT, FIG. 5C is a correspondence relation between secondary counter control codes and counters, and FIG. 5D is a count threshold LUT. is there.

6A and 6B are diagrams showing the operation range of each counter in the counter circuit.

FIG. 7 is a flowchart showing a comparison procedure of a comparison circuit.

8 is a block diagram of an example of the encoding device shown in FIG.

9 is an example of the quantization LUT shown in FIG.

FIG. 10 is an example of a final quantization state.

11 is a flowchart showing an AC component encoding procedure by the encoding device shown in FIG.

12A and 12B are examples of the two code LUTs of FIG.

13A and 13B are diagrams showing a data read order in the data read circuit of FIG.

14 (a), (b), (c), (d), and (e) are diagrams showing an example of the contents of the code configuration.

[Explanation of symbols]

 1 frame memory 2 read-out device 3 two-dimensional discrete cosine transform device 4 quantizer 5 classification device 6 block buffer memory 7 encoder

Claims (5)

[Claims] 1. A reading means for sequentially reading digitized gradation image data in block image units, and a two-dimensional orthogonal transformation for obtaining transform coefficient data by performing two-dimensional orthogonal transformation on the read block image. Means, a classifying means for calculating a statistic of the transform coefficient data in block units, and classifying the block into a plurality of classes based on the statistic of the calculation result, and the transform coefficient data based on the classification result. Quantizing with a predetermined quantization width, and coding means for coding, and the correspondence between the block and the quantization width assigned to the block is controlled by a parameter given to the classification means. Characteristic gradation image data compression device. 2. At least one block and a quantization width assigned to the block by the classifying unit by modifying the value of a part or all of the data input to the classifying unit by the parameter. The gradation image data compression apparatus according to claim 1, wherein the correspondence between the image data and the image data is controlled. 3. The correspondence between at least one block and the quantization width assigned to the block by the classifying unit is modified by modifying the statistic calculation method of the classifying unit with the parameter. The gradation image data compression device according to claim 1, wherein the compression device is controlled. 4. The gradation image data compression apparatus according to claim 1, wherein the two-dimensional orthogonal transformation is a two-dimensional discrete transformation. 5. The parameter corrects the correspondence between the class assigned to each of the blocks by the classifying unit and the quantization width assigned to each of the blocks by the classifying unit, whereby at least one of the blocks is corrected. The gradation image data compression apparatus according to claim 1, wherein the correspondence between a block and a quantization width assigned to the block is controlled by the classification unit.