JPH04175067A - Image compression/restoration device - Google Patents

Abstract

PURPOSE:To make image quality constant by means of a compression rate within a restricted range by discriminating whether or not the deterioration of an image indicated by image information transmitted by a quantization part to an encoding part is proper and controlling the compression in a re-quantization part based on the discriminated result. CONSTITUTION:An image compression/restoration device 1 consists of a compression part 2 and a restoration part 3, original image obtained by a photographing device 4 is compressed in the compression part 2 and stored in a file device 40 and will later be restored in the restoration part 3 and supplied to a display device 5. The compression part 2, which consists of a space processing part 6, re-quantization part 7, an image quality evaluation part 8 and an encoding part 9, divides the original image into blocks and executes orthogonal transformation for concentrating energy on a particular frequency and obtains a multilevel gradation. A buffer is provided at the poststage immediately after the orthogonal transformation, the re-quantization which serves to compress the image information is executed and restoration evaluation is executed at the poststage immediately after the re-quantization. This operation is continued to execute until whether or not the deterioration of the image is proper can be obtained, and thus finally the image quality at a certain level can be obtained.

Description

DETAILED DESCRIPTION OF THE INVENTION [Object of the Invention] (Industrial Application Field) The present invention relates to an image compression and decompression device that compresses and decompresses, for example, medical images, and particularly relates to an image quality improvement technique.

(Prior Art) Conventionally, there is an image compression/decompression device called a block adaptive DCT compression method. This block adaptive DCT compression method was aimed at keeping the compression ratio constant. This is because conventional image compression techniques have been developed primarily for use in communications. In other words, the capacity speed of a communication channel is usually determined, and since the capacity speed of the communication channel is quite low, the capacity that can be sent in a certain period of time is limited. In this case, in order to always use the communication channel to the maximum extent possible, it is necessary to keep the amount of data generated in a certain period of time constant.

For example, in an example of the conventional block adaptive DCT compression method, as shown in FIG. 13, after the original image is divided into blocks (step 5T1), orthogonal transformation (typically DCT) that concentrates energy at a specific frequency is performed. is performed to obtain the gray scale of the image information (step 5T2). Then, requantization is performed to compress the image information by applying quantization width and threshold processing to the gray scale of the image information (step 5T3), and after this requantization, the compressed image information is Encode (step 5T4). A buffer (Rate-buffer) is provided immediately after this encoding (step 5T5), and by providing a feedback mechanism that feeds back re-quantization parameters such as threshold and quantization width until the re-quantization step, it is possible to The compressed image was reconstructed based on the encoded image information while keeping the amount of data generated constant. In addition, in this feedback mechanism, when the original image is divided into blocks in step STI as shown in FIG. 14, for example,
If the compression capacity of the block a in the figure is larger than the preset value, the requantization parameters such as the quantization width are changed in step ST5 so that the compression capacity of the block b is reduced.
If the compression capacity of the block a is small, the requantization parameters such as the quantization width are changed in step ST so that the compression capacity of the block b is increased.
By setting a buffer of 5, the compression ratio is set to 15.
It can be made constant as shown in the figure.

(Problem to be Solved by the Invention) However, in the case of an image compression/decompression device applying the conventional block adaptive DCT compression method described above, non-optimization of feedback and non-uniformity of image quality as described below occur. There was a problem that this caused a problem.

<Non-optimization of feedback> By dividing the original image into blocks, the characteristics of the image differ somewhat from block to block. In this case, in the conventional technique, requantization is determined based on the compression capacity of the previous block, rather than compression according to the image characteristics of each block. Parts with a lot of image information often span several blocks. Naturally, the compression capacity should be distributed evenly across these blocks, but if the capacity of one block increases, the next block will be controlled to a lower level. As a result, the amount of stored information differs between blocks, resulting in large differences between blocks and block artifacts in the restored image. Note that this phenomenon cannot be eliminated as long as blocking is performed in principle.

<Non-uniformity of image quality> Image quality and compression rate are mutually exclusive. If you try to increase the compression rate, the image quality will deteriorate, and if you try to increase the image quality, the compression rate will deteriorate. Furthermore, if an attempt is made to keep the compression rate constant, the image quality will vary from block to block image. Conversely, if you try to keep the image quality constant, the compression ratio will fluctuate.

Therefore, conventionally, a constant compression rate does not provide constant image quality, and it has been difficult to maintain image quality above a certain level required for medical images. .

The present invention has been made with attention to such circumstances, and its purpose is to provide an image compression and decompression device that can obtain constant image quality within a limited range of compression ratios.

[Structure of the Invention] (Means for Solving the Problem) In order to achieve the above object, the present invention compresses the image information by applying quantization width and threshold processing to the density gradation of the image information. An image compression/decompression device comprising a requantization unit for encoding compressed image information and an encoding unit for encoding compressed image information,
an evaluation unit that obtains an index for determining whether the degree of image quality deterioration indicated by the compressed image information sent from the requantization unit to the encoding unit is appropriate; each time the evaluation unit makes a negative determination, The present invention is characterized in that a feedback mechanism is formed to control compression in the requantization unit until the evaluation unit makes an appropriate determination.

Further, the image according to the present invention is characterized in that it includes a restoring device that restores the encoded image information, and a display device that displays the image information restored by the restoring device and the index. With the configuration of the compression/decompression device, the feedback mechanism controls whether or not the requantization unit performs compression according to the suitability determination of the evaluation unit, so that a relationship in which the image quality is constant can be obtained. Therefore, it is possible to maintain image quality above a certain level required for medical images.

(Embodiment) FIG. 2 is a block diagram schematically showing an image display system to which the image compression and decompression apparatus of the present invention is applied.

In this image display system, an image compression and restoration device 1 includes a compression section 2 and a restoration section 3, and the compression section 2 compresses an original image obtained by a photographing device 4, stores it in a file device 40, and restores it later. The data is restored by the copy 3 and delivered to the display device 5.

The compression unit 2 includes a spatial processing unit 6. Re-quantization section 71 image image value section 8. The encoder 9 is equipped with an encoding unit 9, and after converting the original image into blocks (step 101), as shown in FIG. (step 102).

Then, a buffer is provided immediately after the orthogonal transformation (
Step 103), the gradation of the image information is subjected to quantization width and threshold processing through this buffer, and requantization is performed to compress the image information (step 104). Immediately after this requantization, restoration evaluation is performed (step 105). That is, in step 105, it is determined whether or not the degree of deterioration of the image compression restoration performed in step 104 is appropriate based on the preset image quality level. If it is determined to be appropriate, the process proceeds to the next step 106, and the compressed image is encoded. If the determination is negative, the parameters indicating the compression conditions are changed again, and the requantization process in step 104 is executed again, and then in step 105, the image quality is determined again. Repeat this until a suitable judgment is obtained. This makes it possible to maintain image quality above a certain level.

Next, the image quality evaluation section 8, which is the center of the present invention, will be described. Now, examples of indicators used to judge the image quality include: 1) SN ratio between the original image and the compressed and decompressed image 2) Block artifact index of the compressed and decompressed image (described later) 3) Indices for each image quality element of the compressed and decompressed image and their values Comprehensive evaluation 4) Includes spectrum analysis before and after requantization.

First, the image quality evaluation section of the first embodiment corresponding to the above index 1) will be described. The image quality evaluation section of this first embodiment includes an SN ratio calculation section 1 as an image quality index calculation section as shown in FIG.
Has 0. This SN ratio calculation unit 10 is an image restoration circuit that receives requantized data 11a from an internal buffer 11 that temporarily stores original image data and requantized data, and performs various processes using an inverse quantization function 12a and an inverse DCT function 12b. 12
, a difference image creation circuit 13 that receives the output data of the image restoration circuit 12 and the original image data llb from the internal buffer 11 and creates a difference image; SN ratio calculation circuit 1 that calculates the SN ratio by receiving the original image data lla of
4.

The image quality level determination circuit 16 compares and determines the value of the SN ratio calculation section 10 with the image quality reference value 15a preset in the storage section 15. When the judgment result is appropriate, a transfer command to the encoding unit 9 is sent to the internal buffer 11, and the encoding unit 9 encodes it (step 106 in FIG. 1).
Have them do it.

When a negative determination is made, the requantization parameter 15b performed this time is imported from the storage unit 15 into the parameter setting circuit 17 by restarting the requantization parameter setting circuit 17, and further optimization is performed based on the requantization parameter 15b. A requantization parameter 17a is calculated and set to the requantization unit 7, and at the same time, the requantization parameter value setting circuit 17 is set to the preprocessing unit 6.
By sending a reprocessing instruction to the buffer of
A requantization unit 7 is configured to perform reprocessing.

Note that the image restoration circuit 12 in the image quality evaluation section of this embodiment
Based on the output data, the restored image can be displayed on a display device 18 different from the display bag [5 in FIG. 2]. Further, an input device 20 is attached to the image quality level determination circuit 16, and a determination result can be inputted from this input device 20 to the image quality level determination circuit 16 as required.

In the compression method (Chen's 5AC) in this embodiment, the requantization parameters are the step width and the threshold value (th
(reshold value). There is a correlation between these and the SN ratio as shown in FIG. 4, and between the two, the threshold value changes the SN ratio more than the step width. That is, the threshold value is changed when changing the SN ratio sharply, and the step width is changed mainly when making fine adjustments. Of course, you can change them at the same time.

Furthermore, control based on the compression ratio may be added. Although the image quality (in this example, SN ratio) is better, there is a reciprocal relationship between image quality and compression rate as shown in Figure 5, and there are various parameters to achieve the same compression rate. , the range of optimal control is limited. Therefore, with certain set parameters, even if the image quality is exceptionally good, the compression ratio may be poor. It goes without saying that the initially set parameters are values that are likely to yield the desired compression ratio.

Therefore, in the second embodiment, the image quality evaluation section includes a compressed data amount calculation section 21 and a compression ratio level determination circuit 2, as shown in FIG.
2 is provided, and a compression ratio of a satisfactory level within the allowable image quality level can be obtained in the same manner as the image quality level determination performed by the combination of the image quality index calculation unit 23 and the image quality level determination circuit 24. These determination results are then passed to the optimal compression determination circuit 25, and the requantization parameter value setting circuit 17 is restarted based on the optimal compression determination circuit 25. In the figure, 11 is an internal buffer, and 15 is a storage unit, which has the functions described in FIG. 3. This makes it possible to achieve optimal control in the image compression device.

Here, since the compressed data amount calculation unit 21 calculates the compression rate before encoding is performed, the compression rate is not accurate, but
Calculate the entropy of the data and get an almost accurate value. Alternatively, the compressed data amount calculation unit 21 and the compression rate level determination circuit 22 may be placed after the encoding in step 106 in FIG. 1 instead of after the requantization in step 104 in FIG. 1, and an accurate compression rate can be used. good. This is a conventional example (Chen's 5A
This is nothing but the same configuration as C).

Furthermore, although there is generally a fairly strong correlation between the SN ratio and image quality (original image quality), they are not exactly equivalent. According to an experiment conducted when completing this embodiment, a correlation is recognized for roughness up to 3 dB, but a difference of 1 dB is not significant in terms of image quality. In fact, the opposite may be true. SN ratio is 1dB
In some cases, the worse the image quality, the better the image quality. Therefore, at this time, it is also possible to have the user directly view the restored image and leave the decision to the user. Specifically, in FIG. 3, the output data of the image restoration circuit 12, that is, the restored image, is displayed on the display device 19, and if necessary, the differential image creation circuit 13 is displayed.
It is also possible to display the output data and have the user input the judgment of suitability or disapproval using the input device 20 such as a keyboard.

Of course, at this time, from the judgment of the value of the SN non-calculation unit 10,
Needless to say, the user's judgment takes precedence. Furthermore, it is also possible to display the value of this SN non-calculation unit 10 on the display device 19 as a reference.

In each of the embodiments described above, the SN ratio was used to calculate the image quality index, but a block artifact index can be used instead of the SN ratio. FIG. 7 shows the functional block configuration of an embodiment using the block artifact index in 2) above, in which a block artifact calculation section 26 is provided in place of the SN ratio calculation section used in each of the embodiments described above. .

The block artifact calculation unit 26 uses the image restoration circuit 12 to calculate the original image data lla from the internal buffer 11.
Performs image restoration processing, then calculates artifacts in the X direction of the restored image by differential processing in the differential processing circuit 27, calculates artifacts in the y direction of the restored image by cumulative processing in the absolute value cumulative processing circuit 28, and calculates peaks. Calculating the averaged peak value by the value average calculation circuit 29 and passing it to the block artifact calculation circuit 31;
The background average calculation circuit 30 calculates a background average value based on the restored image obtained by the image restoration circuit 12 and passes it to the block artifact calculation circuit 31.
The image quality level judgment circuit 16 sends out the heplock artifact index value so that the image quality level judgment can be made.
(See Figure 8). Note that the X direction and the X direction may be reversed, and block artifacts in both directions are usually measured.

Furthermore, FIG. 9 shows the functional configuration of an embodiment corresponding to the index for each image quality element of the compressed and decompressed image and the comprehensive evaluation of them in 3) above.

In this embodiment, the original image data l1g from the internal buffer 11 is subjected to image restoration processing by the image restoration circuit 12, and then the sharpness calculation section 32, graininess calculation section 33, gradation calculation section 34, color reproduction Each process of the quality calculation unit 34 is performed in parallel, and the results of each of these processes are passed to the overall image quality calculation unit 36.
The overall image quality data is obtained and sent to the image quality level determination circuit 16, so that the image quality level determination circuit 16 can determine the image quality level. Note that, instead of determining the overall image quality, one of the main factors among each pixel may be used as a representative.

Incidentally, indicators for each of these image quality elements are specifically shown, for example, in Japanese Patent Application No. 61-26766 filed by the present applicant.

From another perspective, the SN ratio is nothing but the simplest index for calculating gradation.

In each of the embodiments described above, an image restoration circuit is provided in the image evaluation unit to calculate the restored image quality, but these are disadvantageous in terms of shortening the processing time for calculating the restored image quality. Therefore, considering that the processing in the coefficient space determines the image quality, it is possible to evaluate the image quality by comparing the coefficient space before and after compression processing, and to shorten the processing time for calculating the restored image quality. good. For example, (i) the sum of the coefficients (power) included in a certain low frequency region or its totalization (11) the sum ratio of cut high frequency components before and after requantization, and its waveform (is it a characteristic shape? (iii) For example, the sum of high frequency components or their totalization (the sum of requantized coefficients or their overall ratio) before and after requantization.

Among them, the functional configuration of one embodiment corresponding to the above (11) is shown in FIG. 10, and will be explained with reference to FIG. 11. 11(a) from the buffer of the preprocessing section 6 and FIG. 11(b) from the internal buffer 11.
) is received by the difference coefficient multiplication circuit 37, the difference coefficient multiplication circuit 37 obtains the characteristic data as shown in FIG. Then, this recognition result is sent to the image quality level determining circuit 16, so that the image quality level determining circuit 16 can judge the image quality level.

Here, to explain the basis for (i), (ii), and (iii) above, in (i) above, even if the coefficient value is small, low frequency components are important for image quality, and blurring and
Since this leads to the creation of mosaic patterns, etc., it means the degree to which they are preserved, and is intended to represent the image quality.

In (ii) above, the cut of the high frequency component appears as a professional artifact, so the degree of cutting is the degree of the cut.

The above (iii) is the degree of conservation of the overall coefficients.

These are used in conjunction with the requantization method (compression method) and those that are effective for the determination. For example, in the embodiment shown in FIG. 10, threshold is applied, but since the effect of rounding due to the step width has a strong influence on image quality,
The threshold performance and the step width may be separately weighted and determined. The threshold is
Furthermore, the completely cut component is a problem, and it is also related to its frequency. For example, by using the characteristic data as shown in the w412 diagram (figure a) and the characteristic data as shown in the same figure (b), the same figure ( C), the appropriate frequency weighting is summed (
The total power) may be calculated and this value may be used as an image quality index.

As mentioned above, the processing at the time of compression has been described for each embodiment of the present invention, but at the time of decompression, the decompression may be performed using the processing steps shown in FIG. In addition, in any of the embodiments of the present invention, processing time is required for image compression or restoration than in the past, but this does not cause any inconvenience in practice for the following reasons.

Users who are doctors or technicians do not photograph patients using X-ray imaging equipment, CT equipment, MHI equipment, endoscope equipment, ultrasound diagnostic equipment, etc. over a 24-hour period; I am currently filming intermittently. Even for the same patient, one body position and dose.

Different methods are used, and even in videos like DF, the pause time between patients is relatively long. Also,
The shooting time is rarely night, but usually during the day. Therefore, the compression according to the present invention may be performed during non-photographing hours at night. Naturally, this is done as a pack-ground process and has nothing to do with the user. Of course, it can also be placed under the user's knowledge.

Note that although each of the embodiments of the present invention described above takes SAC as an example, the present invention is not limited thereto; for example, even if other requantization methods are used, image quality can be improved by parameter control. Any method of operation is included in the present invention. In addition, although the above-described embodiments of the present invention have been described with reference to block orthogonal transform, this includes not only full orthogonal transform, but also linear predictive coding in which the orthogonal transform section is replaced with a prediction section, and similar block coding. It can also be applied to Furthermore, vector quantization coding in which vector quantization (VQ) is performed in a requantization unit without a blocking unit/conversion unit, and D
The present invention can also be applied to hybrid compression methods such as CT-VQ. Furthermore, the image quality evaluation department uses AI technology, image recognition,
With the development of cognitive science and technology, it goes without saying that the present invention can be applied to technologies that have knowledge databases and have intelligent functions.

[Effects of the Invention] As explained above, the present invention provides requantization in accordance with the determination result of the evaluation unit that determines whether or not the degree of image quality deterioration indicated by the image information sent from the requantization unit to the encoding unit is appropriate. Since it forms a feedback mechanism that controls compression in the converter, it is possible to maintain constant image quality within a limited range of compression ratios.

[Brief explanation of drawings]

FIG. 1 is a flowchart showing a process in a main part of an image compression and decompression device to which the present invention is applied, FIG. 2 is a block diagram showing an outline of an image display system to which the image compression and decompression device of the present invention is applied, and FIG. The figure is a block diagram showing the functional configuration of the first embodiment of the image quality evaluation section, FIG. 4 is a diagram showing the correlation between the threshold value or step width and the SN ratio, FIG. 6 is a block diagram showing the functional configuration of the second embodiment of the image quality evaluation section, FIG. 7 is a block diagram showing the functional configuration of the third embodiment of the image quality evaluation section, and FIG. 8 is a waveform diagram showing the image quality level judgment situation. Figures 9 and 10 are block diagrams of the functional configuration of other embodiments of the image quality evaluation section, respectively.
12 is a diagram showing the process of determining the image quality level, FIG. 12 is a diagram showing the process of determining the image quality index, and FIG. 13 is a flowchart showing the process in the main part of a conventional image compression and decompression device.
FIG. 14 is a diagram schematically showing how the original image is divided into blocks, and FIG. 15 is a diagram showing waveforms how the compression ratio is made uniform. 1... Image compression/decompression device 2... Compression unit 3... Restoration unit 4... Photographing device 5... Display device 6... Preprocessing unit 7... Requantization unit 8... Image quality evaluation section 9... Encoding department agent Patent attorney Nori Ken Tasuku Kondo Agent Patent attorney Takeshi Kondo Go to Figure 1 aR 2 Degree setting @ 4 Figure 5 Cause @ 6 Figure 7 @ Figure 8 Figure 10 Figure 13 Figure 15 i - $ bottle 9

Claims (2)

[Claims] (1) A requantization unit that compresses the image information by applying quantization width and threshold processing to the density gradation of the image information, and an encoding unit that encodes the compressed image information. The image compression and decompression device includes an evaluation unit that obtains an index for determining whether or not the degree of image quality deterioration indicated by the compressed image information sent from the requantization unit to the encoding unit is appropriate, and the evaluation unit An image compression and decompression device comprising: a feedback mechanism that controls compression in the requantization unit every time the evaluation unit makes a negative judgment until the evaluation unit makes an appropriate judgment. (2) Image compression according to claim 1, further comprising a restoring device that restores the encoded image information, and a display device that displays the image information restored by the restoring device and the index. Restoration device.