JPH0923426A - Method and device for decoding picture - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a correction picture of high picture quality with the absolute minimum number of circuit constitution in picture decoding. SOLUTION: When decoding is continuously executed on a real time, switches 116 and 107 are changed over to contact points 1162 and 1071 . A low pass filter processing, an orthogonal conversion processing, a clipping processing and an inverse orthogonal conversion processing are executed on a decoding picture 106, and a corrected picture 113 is obtained. Then, it is outputted to an output terminal 114 and is accumulated in a frame memory 115. When decoding is stopped in a prescribed frame, the switch 107 is changed over to a contact point 1072 and a series of processings are repetitively executed on the corrected picture 113 accumulated in the frame memory 115.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention divides a digital image signal into small blocks, and quantizes an orthogonal transform coefficient obtained by performing an orthogonal transform for each small block with a predetermined quantization step width. The decoded representative value is obtained, the code corresponding to the quantized representative value is allocated, the recorded code is decoded to reproduce the quantized representative value, and the reproduced quantized representative value is subjected to inverse orthogonal transform to obtain a decoded image. The present invention relates to an image decoding method and apparatus for obtaining a signal.

[0002]

2. Description of the Related Art As a highly efficient image coding method, there is a method using an orthogonal transform represented by a discrete cosine transform.

FIG. 4 is a block diagram of a general orthogonal transform coding / decoding device. First, the image 202 input from the input terminal 201 is N × N in the small block dividing unit 203.
It is divided into small blocks. The orthogonal transformation unit 204 performs orthogonal transformation on each of the divided small blocks, and the obtained orthogonal transformation coefficient is quantized with the quantization characteristic determined by the quantization unit 205. Next, the code allocation unit 206
At, a code is assigned to the quantized representative value, and the encoded data 207 is stored or transmitted through the output terminal 208. On the other hand, at the time of decoding, the code decoding unit 211 decodes the code from the coded data 210 obtained from the input terminal 209, and the inverse quantization unit 212 reproduces the quantized representative value for each orthogonal transform coefficient, and the inverse orthogonal transform is performed. Part 21
Inverse orthogonal transformation is performed at 3, and the decoded image data 214 is output to the output terminal 215.

As shown in FIG. 5, the quantization method is generally such that when the input value F is F _i ≤F <F _{i + 1} , the quantization representative value is F _i '(where F _i ≤F _i It is set so that '<F _{i + 1} ).

In such an orthogonal transform coding method, coding is performed in units of small blocks, and when the compression rate is set high, the quantization is rough, so that the decoded image has block noise or the like. There is a problem that the deterioration of the image quality is detected.

Various methods have been considered to avoid this, but a plurality of constraint conditions that the original image must satisfy are set so that the decoded image deteriorated by encoding satisfies all of these constraint conditions. There is a method to reconstruct the image as close as possible to the original image as a result (A. Zakhor, “Iterative procedures for reductionof
blocking effects in transform image coding, ”IEEE
Trans. On Circuitsand Systems for Video Technolog
y, Vol.2, No.1, pp.91-95, March 1992).

In the method in this document, the constraint condition is
The condition that the original image is smooth and the condition that the range in which the true value before quantization is present are limited from the quantized representative value are used. The former is achieved by applying a low pulse filter to the image. In the latter case, when the quantized representative value is F _i ′, the true value F before quantization is F _i
Since it must exist in ≦ F <F _{i + 1} , it may be modified to satisfy this.

Specifically, first, a low pass filter is applied to the obtained decoded signal. Next, the low-pass filtered signal is orthogonally transformed. The orthogonal transform coefficient F _L obtained at this time is different from the coefficient F _i ′ obtained by orthogonally transforming the signal before the low-pass filter is applied, and does not necessarily satisfy the latter condition. Therefore, _FL is modified according to the following conditions.

[0009]

[Equation 1] That is, if the value exceeds the true value range, the clipping process is performed to the maximum value or the minimum value in the true value range. The orthogonal transform coefficient corrected by clipping is subjected to inverse orthogonal transform to obtain corrected image data. A final corrected image is obtained by repeating the above low-pass filter, orthogonal transformation, clipping, and inverse orthogonal transformation a predetermined number of times.

FIG. 6 shows the method of correcting a decoded image described above.
It is a block diagram to perform. Input terminal 101 when decoding
From the encoded data 102 input from the code decoding unit 10
After the code is deciphered in step 3, the dequantization unit 104 demultiplexes each code
The quantized representative value for the alternating transform coefficient is reproduced, and the inverse orthogonal transform is
Inverse orthogonal transform is performed in the conversion unit 105 and the decoded image data 10
6 is obtained. For the decoded image data 106,
Filter unit 108₁Filtered in
Then, the orthogonal transformation unit 109₁ Orthogonal transformation is performed in
The extracted orthogonal transform coefficient is used as the clipping unit 110.₁ Entered in
You. Clipping part 110₁ In, the inverse quantizer 1
The range setting unit 11 according to the quantized representative value used in 04.
Use the maximum and minimum values that can be taken by the value before quantization obtained by 0
Then, the orthogonal transform coefficient is modified. Modified orthogonal transform
The coefficient is inversely transformed by the inverse orthogonal transformation unit 111, and 1
Third corrected image 113₁ Is obtained. This modified image 11
3 ₁ For the low pulse filter unit 108_Two At 2
The second low-pass filter is applied, and the orthogonal transformation unit 109_Two 
The orthogonal transformation coefficient is applied to the obtained
Ping unit 110_Two Is input to Clipping 110
_Two In the same way as the first time, the orthogonal transformation coefficient is corrected.
Then, the corrected orthogonal transform coefficient is applied to the inverse orthogonal transform unit 111._Two To
Inverse conversion is applied to the second corrected image 113._Two Get
Can be The process is repeated the same number of times
And the obtained n-th corrected image 113_nFinally output
It is output to the terminal 114.

[0011]

In the method as described above, it is necessary to repeat low-pass filter processing, orthogonal transformation, orthogonal transformation coefficient clipping processing, and inverse orthogonal transformation processing for the decoded video signal. It is known that at least several iterations are required to remove block distortions and the like that appear due to orthogonal transform coding, and the image quality improves as the number of iterations increases.

Here, when decoding is required for a still image consisting of only one frame, the circuits for low-pass filter processing, orthogonal transformation, orthogonal transformation coefficient clipping processing and inverse orthogonal transformation processing are required for decoding. One by one, the corrected image can be input to the same circuit again by repeating it one after another.

However, if this method is applied to a moving image, it is necessary to perform the above-described iterative processing for each frame. Therefore, for example, when iterative processing is performed n times for one frame, low-pass processing is performed. N circuits for filter processing, orthogonal transformation, orthogonal transformation coefficient clipping processing, and inverse orthogonal transformation processing are required.

As described above, in order to converge the corrected image finally obtained for the moving image signal to the one close to the original image, there is a problem that a huge amount of hardware is required. .

It is an object of the present invention to provide an image decoding method and apparatus capable of obtaining a corrected image with high image quality with a minimum necessary circuit configuration.

[0016]

According to an image decoding method of the present invention, a process of applying a low-pass filter to an obtained decoded image, and an image signal after the low-pass filter is divided into small blocks and orthogonal transformation is performed. The process, the process of determining whether the obtained orthogonal transform coefficient exists in the specified range, the process of clipping in the range when it does not exist, and the inverse process of the orthogonal transform coefficient subjected to the clipping process. The process of obtaining a corrected image by orthogonal transformation is not executed when the image is decoded in real time, or is executed only once, and when decoding is stopped at a certain frame, it is repeatedly executed while stopped. And

Further, the image decoding apparatus of the present invention has an input terminal, an output terminal, a code decoding unit for decoding the code of the encoded data input from the input terminal, and an orthogonal output from the output of the code decoding unit. The inverse quantization unit for reproducing the quantized representative value for the transform coefficient, the inverse orthogonal transform unit for inversely orthogonally changing the quantized representative value to obtain the decoded image data, and the quantized representative value from the true of the coefficient. A range detection unit for obtaining the maximum value / minimum value of a range of values, a frame memory for storing a corrected image, and the output of the inverse orthogonal transformation unit, one contact or the other contact connected to the output terminal And a first switch for switching and outputting to the other contact, and in the case of performing the code continuously in real time, it is switched to the output of the inverse orthogonal transforming unit via the first switch. , Decryption When stopped at a certain frame, a second switch connected to the output of the frame memory, and a low-pass filter unit for inputting decoded image data or a corrected image through the second switch and performing filtering. An orthogonal transform unit that obtains an orthogonal transform coefficient by orthogonally transforming the output of the low-pass filter unit, and a clipping unit that corrects the orthogonal transform coefficient so as to be between the maximum value and the minimum value obtained by the range detection unit. And an inverse orthogonal transform unit that performs inverse transform on the corrected orthogonal transform coefficient to obtain a corrected image, outputs the corrected image to the output terminal, and stores the corrected image in the frame memory.

[0018]

BEST MODE FOR CARRYING OUT THE INVENTION During normal decoding / reproduction, that is, when an image is being decoded in real time, a series of processes of low-pass filter, orthogonal transform, orthogonal transform coefficient clipping, and inverse orthogonal transform is not executed or Although it will be executed only once, in the state where the coding noise that observes the frame of the coded and decoded video still is still noticeable, the low pass filter, the orthogonal transform, the orthogonal transform coefficient clipping, and the inverse transform coefficient A series of processes of orthogonal transformation will be repeatedly executed many times.

[0019]

Next, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram of an image decoding apparatus according to an embodiment of the present invention. The image decoding apparatus according to the present embodiment includes an input terminal 101, an output terminal 114, a code decoding unit 103 that decodes the code of the encoded data 102 input from the input terminal 101, and an orthogonal output from the output of the code decoding unit 103. Inverse quantization section 104 for reproducing the quantized representative value for the transform coefficient
And the quantized representative value is subjected to inverse orthogonal transformation to obtain the decoded image data 10
6, an inverse orthogonal transformation unit 105, a range detection unit 111 that obtains the maximum and minimum values of the range that the true value of the coefficient can take from the quantized representative value, and a frame memory 115 that stores a corrected image 113. And the output 106 of the inverse orthogonal transformation unit 105 is connected to the output terminal 114 by one contact 1
16 ₁ or the switch 116 for switching and outputting to the other contact 116 ₂ and the switch 116 is the other contact 116 ₂
When the decoding is continuously performed in real time, the switch 107 is connected to the contact 107 ₁ , and when the decoding is stopped at a certain frame, the switch 107 is switched to the contact 107 ₂ and connected to the output of the frame memory 115. Then, the decoded image data 106 or the corrected image 113 is input via the switch 107, the low-pass filter unit 108 that performs filtering, and the orthogonal transform unit 109 that orthogonally transforms the output of the low-pass filter unit 108 to obtain an orthogonal transform coefficient.
And a clipping unit 1 for correcting the orthogonal transform coefficient so as to be between the maximum value and the minimum value obtained by the range detection unit 111.
10 and an inverse orthogonal transformation unit 112 that performs inverse transformation on the corrected orthogonal transformation coefficient to obtain a corrected image 113, outputs it to the output terminal 114, and stores it in the frame memory 115.

Next, the operation of this embodiment will be described.

At the time of decoding, after the code is decoded by the code decoding unit 103 from the coded data 102 input from the input terminal 101, the inverse quantization unit 104 reproduces the quantized representative value for each orthogonal transform coefficient. Then, the inverse orthogonal transform unit 105 performs inverse orthogonal transform to obtain decoded image data 106.

When the decoding is continuously performed in real time, the switch 116 is switched to the contact 116 ₂ and the switch 107 is switched to the contact 107 ₁ , and the decoded image data 106 is input to the low pass filter section 108. . That is, the decoded image data 106 is filtered by the low-pass filter unit 108, then orthogonally transformed by the orthogonal transformation unit 109, and the obtained orthogonal transformation coefficient is input to the clipping unit 110. On the other hand, from the quantized representative value obtained by the inverse quantization unit 104, the range detection unit 111 obtains the maximum value / minimum value of the range that the true value of the coefficient (value before quantization) can take, and the clipping unit 110. Entered in. Clipping part 110
In, the orthogonal transform coefficient is modified so as to fall between the obtained maximum value and minimum value. The corrected orthogonal transform coefficient is inversely transformed by the inverse orthogonal transform unit 112 to obtain a corrected image 113, which is output to the output terminal 114 and input to the frame memory 115.

On the other hand, when the decoding is stopped at a certain frame, the switch 107 is switched to the contact 107 ₂ (the switch 106 remains switched to the contact 116 ₂ ) and stored in the frame memory 115 in the low-pass filter section 108. The corrected image 113 that has been input is input. That is, the corrected image data 113 is filtered by the low-pass filter unit 108, then orthogonally transformed by the orthogonal transformation unit 109, and the obtained orthogonal transformation coefficient is input to the clipping unit 110. On the other hand, from the quantized representative value obtained by the inverse quantization unit 104, the maximum value / minimum value of the range in which the true value (value before quantization) of the coefficient can be obtained by the range detection unit 111, and the clipping unit 110 Entered in. Clipping part 110
In, the orthogonal transform coefficient is modified so as to fall between the obtained maximum value and minimum value. The corrected orthogonal transform coefficient is inversely transformed in the inverse orthogonal transform unit 112 to obtain a new corrected image 113, which is output to the output terminal 114 and stored in the frame memory 115. The contents of the frame memory 115 are updated every time a loop of low-pass filter processing, orthogonal transformation, orthogonal transformation coefficient clipping processing, and inverse orthogonal transformation processing is performed. And
While the decoding is stopped, these processes are repeatedly executed, and the output terminal 114 has the inverse orthogonal transform unit 112 at that time.
The image data 113 processed in is output.

When the series of processes are not executed at the time of real-time decoding, the switch 116 turns the contact 116.
_It is switched to ₁ , and the decoded image data 106 is output to the output terminal 104 as it is.

Next, a specific example of this embodiment will be shown. In this example, the case of using discrete cosine transform as an information compression method will be considered.

In the encoding, the input image is divided into 8 × 8 small blocks f (i, j) (i, j = 0,1,2, ...
Discrete cosine transform is applied to 7) and transform coefficient F
(I, j) is obtained. The transform coefficient F (i, j) is quantized by a quantizer having a step width of 2d as shown in FIG.

At this time, if the value F (i, j) before quantization is 2nd-d≤F (i, j) <2nd + d, the quantized representative value F '(i, j) is F'. (I, j) = 2nd. In this case, conversely at the time of decoding, F ′ (i, j) =
When 2nd is transmitted, the true value F before quantization
(I, j) is 2nd−d ≦ F (i, j) <2nd + d
It can be seen that it is within the range.

Consider a case where the quantized representative value F ′ (i, j) obtained by the inverse quantization unit 104 at the time of decoding is 2nd. First, F ′ (i, j) is subjected to inverse discrete cosine transform in the inverse orthogonal transform unit 105, and the decoded image 106

[0030]

[Outside 1] Get. This decoded image

[0031]

[Outside 2] First, the low-pass filter unit 108 applies the following two-dimensional low-pass filter.

[0032]

[Equation 2] This low-pass filter is applied across the boundaries of small blocks. The orthogonal transform unit 109 performs a discrete cosine transform for each 8 × 8 small block on the image signal to which the low-pass filter has been applied, and obtains the obtained discrete cosine transform coefficient.

[0033]

[Outside 3] And

[0034]

[Outside 4] On the other hand, the clipping unit 110 performs the following clipping processing.

[0035]

(Equation 3) The state of clipping processing is shown in FIG. The inverse orthogonal cosine transform is performed by the inverse orthogonal transform unit 112 on the transform coefficient subjected to the clipping process, and the first corrected image is obtained.

[0036]

[Outside 5] Get. This corrected image if the decoding is done in real time

[0037]

[Outside 6] Is observed as the final image.

On the other hand, when the decoding is stopped at a certain frame, the above process is repeatedly executed to obtain the n-th corrected image obtained at a certain time.

[0039]

[Outside 7] Is observed as the output image at that time.

[0040]

As described above, according to the present invention, only one circuit for each of low-pass filter processing, orthogonal transformation, orthogonal transformation coefficient clipping processing, and inverse orthogonal transformation processing is provided, and a coded and decoded moving image is obtained. In the state where the coding noise, where the image frame is observed still, is very noticeable, the low-pass filtering process, the orthogonal transform, the orthogonal transform coefficient clipping process, and the inverse orthogonal transform process are repeatedly executed. Therefore (actually, the corrected image will converge if you repeat the series of processing of low-pass filter, orthogonal transformation, orthogonal transformation coefficient clipping, and inverse orthogonal transformation a dozen times). A good corrected image can be obtained, and even when decoding is performed in real time, a low-pass filter, orthogonal transform, orthogonal transform coefficient clipping, Since a series of orthogonal transformation processes is executed once (in this case, it is not possible to repeat it many times, but when real-time decoding is performed, observe frames that constantly change as moving images. Therefore, it is not as noticeable as when stationary, so it is sufficient to execute it once. As a result, with the minimum necessary circuit configuration, it is easy to notice the coding noise of the reproduced image in the frame stopped state. The effect is that the image quality improves with time.

[Brief description of drawings]

FIG. 1 is a block diagram of an image decoding apparatus according to an embodiment of the present invention.

FIG. 2 is a diagram showing a quantization characteristic in the embodiment of FIG.

FIG. 3 is a diagram illustrating an orthogonal transform coefficient clipping process in the embodiment of FIG.

FIG. 4 is a block diagram of a general orthogonal transform encoding / decoding device.

FIG. 5 is a diagram illustrating an orthogonal transform coefficient quantization method used for an orthogonal transform code.

FIG. 6 is a block diagram showing a conventional method for correcting a decoded image by repeating a low-pass filter and orthogonal transform coefficient clipping.

[Explanation of symbols]

101 Input Terminal 102 Input Image 103 Code Decoding Section 104 Inverse Quantization Section 105 Inverse Orthogonal Transformation Section 106 Decoded Image Data 107 Switches 107 ₁ , 107 ₂ Contacts 108, 108 ₁ , 108 ₂ Low Pass Filter Sections 109, 109 ₁ , 109 ₂ Orthogonal Conversion unit 110, 100 ₁ , 110 ₂ Clipping unit 111 Range detection unit 112, 112 ₁ , 112 ₂ Inverse orthogonal conversion unit 113, 113 ₁ , 113 ₂ , 113 _n Corrected image 114 Output terminal 115 Frame memory 116 Switch 116 ₁ , 116 ₂ contacts

Claims (2)

[Claims] 1. A digital image signal is divided into small blocks, and an orthogonal transform coefficient obtained by performing an orthogonal transform for each small block is quantized with a predetermined quantization step width to obtain a quantized representative value. , An image decoding method for allocating a code corresponding to a quantized representative value, decoding the recorded code to reproduce the quantized representative value, and inverse orthogonally transforming the reproduced quantized representative value to obtain a decoded image signal In the above, in the obtained decoded image, a process of applying a low-pass filter, a process of dividing the image signal after the low-pass filter into small blocks and performing an orthogonal transform, and an obtained orthogonal transform coefficient within a predetermined range. The process of determining whether it exists,
If it does not exist, the process of clipping within that range and the process of obtaining the corrected image by performing the inverse orthogonal transform on the orthogonal transform coefficient that has been subjected to the clipping process are not executed when decoding the image in real time or only once. An image decoding method, which is characterized in that, when the image is executed and the decoding is stopped at a certain frame, it is repeatedly executed while the frame is stopped. 2. An input terminal, an output terminal, a code decoding section for decoding the code of coded data input from the input terminal, and a quantization representative value for each orthogonal transform coefficient from the output of the code decoding section. The inverse quantization unit for reproducing, the inverse orthogonal transform unit for inverse orthogonally transforming the quantized representative value to obtain decoded image data, and the maximum value of the range that the true value of the coefficient can take from the quantized representative value. A range detection unit that obtains a minimum value, a frame memory that stores a corrected image, and a first switch that switches and outputs the output of the inverse orthogonal transform unit to one contact or the other contact connected to the output terminal. Then, the first switch is switched to the other contact, and when the decoding is continuously performed in real time, it is switched to the output of the inverse orthogonal transform unit via the first switch and the decoding is stopped at a certain frame. In case Is a second switch connected to the output of the frame memory, a low-pass filter unit for inputting the decoded image data or the corrected image through the second switch and filtering, and an output of the low-pass filter unit. Orthogonal transform unit to obtain an orthogonal transform coefficient, a clipping unit that corrects the orthogonal transform coefficient so as to be between the maximum value and the minimum value obtained by the range detection unit, and the corrected orthogonal transform An image decoding device having an inverse orthogonal transform unit which performs inverse transform on a coefficient to obtain a corrected image, outputs the corrected image to the output terminal, and stores the corrected image in the frame memory.