JPS61135286A - Picture data compressor - Google Patents

Abstract

PURPOSE:To apply data compression with a compression factor corresponding to a picture quality requested in the unit of picture by providing a means for setting a factor round-off threshold value and a factor quantized step width corresponding to the set picture quality. CONSTITUTION:Picture data stored in a picture memory 1-14 is read in the unit of blocks for 2-dimensional discrete cosine conversion by a block data reader 1-15. The 2-dimensional discrete cosine converter 1-16 multiplies a 2-dimension discrete cosine conversion matrix and its transposition matrix from the read block data to obtain a conversion factor matrix. A picture quality selector 1-22 outputs the factor round-off threshold value 1-11 and the factor quantized step width 1-12 according to the picture quality switching signal selected by the user corresponding to the requested picture quality. A code data 1-21 formed by a coder 1-19 is added with the threshold value 1-11 and the step width 1-12 by a parameter adder 1-23 and the result is outputted as compressed data 1-13.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to an image data compression apparatus system that allows selection of image quality.

(Prior Art and its Problems) There are many types of images used in the medical field, both X-ray and CT images, which have conventionally been stored in the form of film. However, in hospitals, etc., the number of sheets generated per day is large, and it is necessary to store them for several years, resulting in a huge amount, requiring space for storage and manpower for searching.

Time is a big issue.

In recent years, with advances in digital image processing technology and device technology, it has become possible to digitize medical images for storage and retrieval.

In medical images, high accuracy is required for both the resolution and the number of gradations, at least 1024 x 1024 x 8
It is a bit, no matter how advanced the memory element is.

The problem is the amount of data. For this reason, data compression technology is often used to reduce the amount of stored data.Data compression technology can be broadly divided into predictive coding and conversion coding, but when performing high compression with a compression rate of about 5 or more. Conversion codes are said to be advantageous for

Medical images have strict requirements for one image quality deterioration, and if the signal-to-noise ratio (hereinafter referred to as 87N ratio) becomes too low, the findings will become blurred, causing a practical problem. In conventional image data compression devices, the compression method was fixed, so the Sl.
It was necessary to set a compression parameter that would provide an N ratio, and as a result, a sufficient compression ratio could not be obtained. However, depending on the content of medical images, some require a high 87N ratio, while others are satisfied with a certain 8/N ratio.
For example, in the case of X-ray images, images that require detailed examination because the affected area has not yet been determined require a very high S/N ratio in order to fully reproduce the details. Images of the affected area are clearly visible and should be saved for future reference, and X-ray images of the diseased area are unlikely to be referenced in the future, but should be saved for several years. etc., a very high 87N ratio is not required, but a high compression ratio is required.

(Object of the Invention) An object of the present invention is to provide a variable image quality data compression/expansion system that allows image quality to be selected according to user requirements.

(Structure of the Invention) The image data pressure m device of the present invention includes means for setting the coefficient truncation threshold and coefficient quantization step width corresponding to the image quality by the user setting the image quality as necessary. , means for reading an original image from a storage device in blocks; means for performing orthogonal transformation on the blocks to obtain transform coefficients; , a coefficient truncation means for dividing the coefficient truncation threshold from the absolute value of the conversion coefficient that takes an absolute value not smaller than the coefficient truncation threshold, and quantization of the output of the number truncation means by the coefficient doji conversion step width. means for creating encoded data by assigning a preset code to the converted coefficients, and transmitting the encoded data with the coefficient truncation threshold and the coefficient quantization step width added thereto. and a means for outputting to a storage device or a storage device.

(Example) The present invention will be described in detail below with reference to the drawings.

Here, we mainly focus on X-ray images.

Although the two-dimensional screed cosine transform will be described as an orthogonal transform, it is clear that both CT images may be used as the object, or orthogonal transforms such as Hadamard transform may be used.

FIG. 1 is a block diagram showing an embodiment of the variable image quality data compression device of the present invention.

The image data stored in the image memo IJ-1-14 is read out in block units by the block data reader 1-15, which performs two-dimensional discrete cosine transformation. L
The block data read out after being divided into blocks of 6 pixels and L6 pixels horizontally is subjected to two-dimensional discrete cosine transformation giiFl-16, and then multiplied by the two-dimensional discrete cosine transformation matrix and its transposed matrix. Since one or more calculations that yield a transform coefficient matrix are performed with a finite word length, the calculation results do not correspond to the required image quality in the 0 image quality selection fs1-22, which has already been quantized by rounding. According to the image quality switching signal selected by the user, the coefficient truncation threshold (hereinafter referred to as threshold) set to ``-~-'' of the dynamic range of the conversion coefficient and 8160' of the dynamic range of the conversion coefficient is set to 8160. The combination with the coefficient quantization step width (hereinafter referred to as step width) is output as the threshold t-ii and the step width 1-12.

The transform coefficient matrix obtained by the two-dimensional discrete cosine transformer 1-16 is set to a threshold value 1 in the coefficient truncation 51-17.
-11. According to the comparison result, conversion coefficients whose absolute value is smaller than the threshold value are rounded to zero, and for other conversion coefficients, the threshold value is subtracted from their absolute value.However, the above processing is not performed for the DC component, and the two-dimensional The output of the discrete cosine transformer 1-[6 is output as it is. The output of the zero-order coefficient truncation at-17 is the coefficient quantizer 1.
-18, uniform quantization is performed with a step width of 1-12.

The uniformly quantized transform coefficients are assigned variable length codes preset in accordance with the frequency distribution of the transform coefficients in an encoder 19. The encoded data 1-21 created by the encoder is passed through the parameter adder 1-23 to the threshold value 1-11.
and a step width of 1-12 are added and output as compressed data 1-13 to a transmission line or storage device.

Figure 2 shows a block diagram of the professional data series generator.
A clock is applied to the terminal 4 from a clock generator (not shown) to drive the 1-row address counter 26 and the column address counter n. The output of each address counter is outputted to the terminals n and 1, and the image data of the corresponding address is successively outputted from the one image memory 14, and the image data is written from the terminal U to the block data memo IJ-28. The written image data is output to the two-dimensional discrete cosine transformer via terminal 4.

FIG. 3 shows a block diagram of a two-dimensional discrete cosine transformer. The image signal is divided into small regions as a matrix P
Expressing this as
, k) components from the serial output terminal 32, and (i, j
) components are read by inputting an address from an address generator (not shown) from the terminal 31.

The product of both is also determined by a multiplier 35A. Add this result to the output of register 37A using 36 adders, and store it again in 37 registers.If you fix i and k and repeat j from 1 to n, you will get the product of A and P. of a certain matrix (
i, k) components are found.

The value is stored in the address corresponding to (i, k) of the block data memoria. However, n is the size of the block, and when j starts from 1, the contents of register r are cleared.By repeating the operation of 0 or more from 1 to n for i and P, the matrix of the product of people and P is created. Stored in block data memoria.

Next, in exactly the same manner as above, find the product of person P and input -1. That is, the (i, j) components of A and P stored in the block data memoria and the read-only memo IJ-34
The (k, j) component of the transformation matrix is inputted from the terminal 31 and read out from the address generator (not shown) to calculate the product for j from 1 to n.
Find the (i, k) components of the person PA'', repeat this process, and output the transformation coefficient matrix to the terminal O.Here, A and the person are created by changing the reading method of the exactly same read-only memoria. .Also, since the calculation is performed with a finite word length, the output has already been quantized.

FIG. 4 shows a block diagram of the coefficient truncation device. The conversion coefficients input from the terminal 41 are sent to the absolute value circuit 44 together with vAK supplied from the image quality selector 1-22 via the terminal 42.
Absolute values are taken at A and 44B, and their magnitudes are compared by a comparator 45. At the same time, a threshold value is subtracted from the absolute value of the conversion coefficient in subtraction 1146, the sign of the conversion coefficient is determined by polarity determiner 4, and the product of the two is obtained in multiplication a47 to reproduce the sign of the conversion coefficient. Then, as a result of the comparison 1s49, if the absolute value of the conversion coefficient is smaller than the threshold value, the AND gate 49 is closed and zero is output to the terminal A, and if not, the previous multiplication a
The output of iF47 is output to terminal 4. Also, when the conversion coefficient corresponds to a DC component, the absolute value output of the threshold value is set to zero and 2
The output of the dimensional discrete cosine transformer is directly output to the terminal.

The above operation can be expressed as follows.

However, X is a transformation coefficient, t is a threshold value esgnC.) is a function for extracting the sign, titJ is the row and column number of the transformation coefficient matrix.

FIG. 6 shows a block diagram of the quantizer. The truncated conversion coefficients input from the terminal 61 are divided by the doji conversion step width input from the image quality selector 1-22 via the terminal 62. Uniform quantization is performed in the calculator and output to the terminal section. However, here too, in the case of a DC component, the doji conversion step width is set to 1.

What is output is what is essentially not processed.

Doji conversion step width 1, 2.4. ...and in the form of a power of 2, it is division! 64 only needs to be shifted.

FIG. 5 shows a diagram in which the relationship between the two is plotted using a typical X-ray image as an example, with entropy on the horizontal axis, S/N) f, and various threshold values on the vertical axis.

(The step width is fixed) As a result of considering various values for the step width, we found that the @ value is 1 of the dynamic range of the conversion coefficient.
In the range of /8□6° to 84□6°, 8/N and entropy maintain a linear relationship.

If the value is higher than that, the entropy will not decrease even though the S/N will decrease, and the
If a S/N ratio of dBii or higher is not ensured, delicate affected areas will become blurred, and ink has been confirmed to cause problems. Therefore, it is appropriate to set the threshold value for coefficient truncation at 14□66° to −□6°, which is the dynamic range of the transform coefficient.

Figure 7 shows the relationship between entropy and 8/N.

The threshold value is shown as a parameter, and here the threshold value is fixed, but as mentioned earlier, the relationship between 87N and entropy is maintained in a linear range after ensuring that 37N is greater than d\B. As a result of considering various threshold values, the step width was determined to be 4□6 of the dynamic range of the conversion coefficient.
It has been confirmed that selecting 0 N/8□6° is appropriate.

Figure 1O shows an example of the average values of the 87N ratio and compression rate for multiple X6 images for three combinations of threshold value and step width. 816G) (5W).

-) (-2L → 8160
8160 9 816G possible. In this case, 49
When image quality of about dB is required, the threshold value is 1'-11.
, as a combination of step widths 1-12 '8160 #
8160) and an image quality of about 44 dB is required. 8160), and if an image quality of about 41 dB is acceptable, it outputs (8160'π'').In this way, the image quality selector sets the threshold value as required.

By switching the step width, the compression rate can be increased for images that do not require image quality as high as 1%.

Next, the transform coefficients that have been truncated and quantized as described above are subjected to entropy encoding according to their frequency of occurrence in encoding m1-19. At that time, a variable length Huffman code is simply applied to each component coefficient. In addition to the allocation, taking advantage of the fact that high frequency components within a block are often zero, the coding order of each component is zigzag as shown in Figure 8, so that zeros continue in the latter half. In this way, run-length encoding can be performed for the zero portion, or if there are all zeros from the middle to the end within one block, the encoding can be stopped at that point and an end mark of the block can be sent to improve encoding efficiency. It is possible. A threshold value and a step width are added to the code data created by the code 1S1-19 by a parameter adder to create compressed data 1-13.

An example of the compressed data structure is shown in Fig. 9(a), (bl, and (C)).One entire image consists of a header and the code of each block as shown in Fig. 9(a).The header is The threshold value and step width selected by the image quality selector are stored in the configuration shown in FIG. 9 (bl).
In the configuration shown in Figure (C), non-zero coefficients are assigned Huffman codes, and when zeros continue, their run lengths are stored. If zeros continue from the middle to the end of the block, a block end mark is added immediately after the coefficient sequence without sending a run length of zero.

The compressed data obtained in this way is sent to a transmission path and stored in a file device. In the above embodiment, the threshold value, step, and width selected by the image quality selector are Although we have entered the values as they are, you can also number some combinations and add the number corresponding to the selected combination as a header. In this case, of course, the selection number on the decompressor side is the actual threshold value used.

A conversion circuit is required to convert the step width.

Furthermore, in the embodiment described above, a variable length code that is preset according to the frequency of occurrence is used as the code.

Of course, equal length codes can also be used.

(Effects of the Invention) As described above, the variable image quality data compression device of the present invention supports the image quality required for each image by selecting and using the threshold value and step width corresponding to the required image quality. Data can be compressed and expanded at the same compression rate.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing one embodiment of the present invention. FIG. 2 is a block diagram of a block data serializer. Figure 3 shows a two-dimensional discrete cosine transformer. Figure 4 is a block diagram of the coefficient truncation device, Figure 5 is a diagram showing the relationship between entropy and S/N using the coefficient truncation threshold as a parameter, Figure 6 is a block diagram of the coefficient quantizer, and Figure 7 is a block diagram of the coefficient quantizer.
The figure shows the relationship between entropy and S/N using the coefficient doji conversion step width as a parameter, and FIG. 8 shows the coding order within a block. FIG. 9 shows the structure of the code, and FIG. 10 shows the threshold values. FIG. 3 is a diagram showing the relationship between the set of step and width, SLN ratio, and pressure a ratio. In fig. 1-14... Image memo IJ-, 1-15... Block reader, 1-16... Two-dimensional discrete cosine transformer, 1-17... Coefficient truncation device, 1-18
... Coefficient quantizer-1-19 ... Encoder, 1-2
2... Image quality selector. 1-23... Parameter adder, 26... Row address counter, 27... Column address counter, 28
...Block data memory, 34...Memory for continuous output. That person, 35B...multiplier, 36A, 36B...adder, 37A. 37B...Register, 38...Block data memory. 44A, 44B... Absolute value circuit, 45... Comparator, 46... Adder, 47... Multiplier, 48...
Polarity determiner, 49...AND gate, 64...Divider. Figure 1 Figure 3 Figure 4 Figure 5 Entropy (bit/pel) Figure 6 Figure 7 Figure 8 Figure 9

Claims (1)

[Claims] means for setting a coefficient truncation threshold and coefficient quantization step width corresponding to the set image quality; means for reading the original image from a storage device in blocks; and performing orthogonal transformation on the blocks to obtain transform coefficients. and zeroing the value of the conversion coefficient that takes an absolute value smaller than the coefficient truncation threshold;
Coefficient truncation means for dividing the coefficient truncation threshold from the absolute value of the transform coefficient that takes an absolute value not smaller than the coefficient truncation threshold, and means for quantizing the output of the coefficient truncation means by the coefficient quantization step width. a means for creating encoded data by assigning a preset code to the quantized transform coefficient; and a means for adding the coefficient truncation threshold and the coefficient quantization step width to the encoded data and transmitting the coefficient to the transmission line or An image data compression device characterized by comprising means for outputting to a storage device.