JPS63121375A - Motion compensation interframe coding system - Google Patents

Abstract

PURPOSE:To remove noise in a dynamic area and a static area by controlling the characteristics of a low-pass filter by means of dynamic vector information and the condition of non-coded and coded block, and switching the insertion and non-insertion of a median filter by means of a difference value around the filter. CONSTITUTION:An orthogonal transformer 21 converts the signal of a difference value obtained by means of subtraction at every small block consisting of a prescribed number of picture elements in a subtractor 4 into a coefficient and a quantization part 5 executes quantization. Then, an encoder 6 codes it and an inversion orthogonal transformer 22 converts it into the original picture element. An adder 7 adds a predicative value to it, and a local decoded signal is generated. Thes, a filter control part 11 detects the difference around the median filter 10. If the difference is small, an output from a switch 12 is set to the output from a filter 10, and impulse-like noise lying in the static area. A filter control part 9 controls the characteristics of an LPF8 in accordance wit the condition of a dynamic vector information and non-coding and coding block and removes noise in the dynamic area. Thus, picture quality can be improved.

Description

[Detailed Description of the Invention] [Summary] In a motion compensated interframe coding device, insertion of a filter that removes high frequencies of a decoded image signal is determined based on the size of a motion vector and conditions for determining whether to code or not to code. Alternatively, by controlling based on the difference that occurs before and after inserting the filter, coding noise in the moving domain and pulse-like noise in the static domain are actively removed.

[Industrial application field]

The present invention relates to a motion compensated interframe encoding device for encoding and transmitting images, and in particular, inserting a low-pass filter into a locally decoded signal based on the size of a motion vector and the conditions for encoding or non-coding. This invention relates to motion compensated interframe coding that is adaptively controlled based on errors that occur before and after filter insertion.

Motion compensation coding is a high-efficiency coding method that encodes and transmits moving image information at a low bit rate, and is suitable for effective use of transmission paths and reduction of communication costs.
When performing encoding such as orthogonal transformation, image quality deteriorates due to granular noise and the like.

Inserting a low-pass filter into the decoded output is effective for this purpose, but constant use will result in a significant drop in resolution.

Therefore, there is a need for a method that can reduce noise without reducing resolution.

[Conventional technology]

An example of inserting a low-pass filter in response to the above request is shown in IEICE Technical Report C386-38 (published on July 18, 1986), "A Study of Vector Quantization for Interframe Prediction Error Signals". will be explained with reference to FIG. 1 is a frame memory that stores image information of the previous frame, 2 is a motion vector detection unit, 3 is a variable delay unit, 4 is a subtracter, 3
1 is a vector quantization unit, 32 is a vector quantization decoding unit,
7 is an adder, 8 is a low-pass filter, and 9 is a filter control section.

The image signal of the previous frame stored in the frame memory 1 and the input image signal are compared in a motion vector detection section 2 to generate a motion vector indicating the amount and direction of motion. The variable delay unit 3 changes the amount of delay depending on the vector. The image signal of the previous frame in the frame memory 1 is delayed according to the motion vector in the total delay unit 3 to generate a predicted value signal.

The subtracter 4 subtracts the predicted value signal from the input image signal,
Generates a differential value signal. The vector quantization unit 31 vector quantizes the difference value signal to generate encoded vector error information.

The vector quantization decoding unit 32 decodes the encoded vector error information. The adder 7 adds the decoded error information and the predicted value signal to generate a local decoded output. The local signal output is added to the frame memory 1 as it is or after passing through the low-pass filter 8 and is sent to the frame memory 1 M8. At this time, the low-pass filter 8 is controlled by the filter control section 9 depending on the magnitude of the motion vector. Furthermore, the encoded error information and motion vector information generated in this manner are sent out via a transmission path.

[Problem that the invention seeks to solve]

In the motion-compensated interframe coding method shown in Figure 5, insertion of a low-pass filter for the local decoded output was controlled only by the magnitude of the motion vector, but the prediction error signal was determined to be non-coded. When a low-pass filter is inserted into that block, the image quality gradually deteriorates.

[Means for solving problems]

The present invention solves these problems and further improves the image quality.As shown in the principle structure of FIG. The motion vector detection section 2 compares the image information of the current frame with the image information of the previous frame to detect a motion vector, and the variable delay section 3 generates a delay that changes according to the motion vector. A predicted value signal is generated by applying the signal to the output of the frame memory l, a subtracter 4 subtracts the predicted value signal from the input image signal to generate a difference value signal, and a quantizer 5 generates a difference value signal. is quantized to generate a coded error signal, the decoder 6 decodes the coded error signal, and the adder 7 adds the decoded error signal and the predicted value signal to generate a locally decoded signal. In the motion compensated inter-seam coding method that generates
0, a filter control section 11, and a switch 12 are added.

A low-pass filter 8 and a median filter 10 remove high frequency components and impulse-like components in local decoding.

In addition to the conventional motion vector size, the filter control unit 9
Encoding and non-encoding blocks are determined by observing the quantized signal, and insertion or non-insertion of the low-pass filter 8 is controlled according to the result.

The filter control unit 11 determines the magnitude of the difference between the signals before and after the median filter 10 on a pixel-by-pixel basis, and based on the result, the switch 12 controls insertion or non-insertion of the median filter 10.

[Effect]

By accumulating the locally decoded signal in the frame memory 1 and reading it out, an image signal of one frame before is generated, and a predicted value signal for generating a difference value signal is created from this. is accompanied by quantization noise caused by encoding the difference value signal, so the encoded error signal output also contains noise.

In order to remove noise without reducing the resolution of the still area, a low-pass filter 8 is provided at the local decoded signal output, and depending on the magnitude of the motion vector detected by the motion vector detection means 2, the low-pass filter is applied to the moving area where the motion vector is large. 8 is inserted, and the low-pass filter 8 is controlled not to be inserted in static areas where the motion vector is small. However, if this alone is used for the non-encoded blocks, the image quality will gradually deteriorate.

Therefore, by adding a condition to the filter control unit 9 that determines whether the block is non-encoded or encoded based on the signal from the quantizer 5 and inserts the low-pass filter 8 only into the encoded block, the image quality can be reduced without deterioration. It becomes possible to remove noise occurring in a moving region without reducing the resolution of the region.

Also, the filter control unit 11 detects the difference before and after the median filter 10, and only when the difference is small, the switch 12
By using the output of the median filter 10 as the output of the median filter 10, it becomes possible to remove impulse-like noise existing in a static region.

〔Example〕

FIG. 2 shows one embodiment of the present invention.
The same parts as in the figures are designated by the same numbers and their functions are also similar. Further, 21 is an orthogonal transformer, and 22 is an inverse orthogonal transformer.

In FIG. 2, the difference value signal obtained by subtracting each small block consisting of a certain number of pixels in the subtracter 4 is converted from pixels to coefficients by an orthogonal transformer 21, and then quantized by a quantizer 5. be converted into The quantized coefficients are decoded by a decoding unit 6, converted into original pixels by an inverse orthogonal transformer 22, and then added to the predicted value by an adder 7 to generate a locally decoded signal.

Next, a signal that passes through the median filter 10 and a signal that does not are generated and selected by a switch 12.

Figure 3 is a diagram showing the configuration of the median filter.
(1) is an example of a median filter. Here, the output pixel a' is the median value of the input pixels at b, c, d, and e. Figure 3 (2) is an example of the configuration of (1), and in this figure the median filter 10 is the comparator 41 +
, ROM4 lx 1 selector 413.

First, input pixels a, b, c, d, and e are input to the comparator 41.
Thus, a comparison of 10 and 1 is made.

The comparison result is input to ROM41z and the median value is detected.
The selector 41a is controlled by the result, and the median value a', which is the output value of the median filter 10, is output.

The filter control unit 11 consists of a subtracter 414 and a comparator 416 shown in FIG. 3(2), and the output value a of the median filter IO and a signal not subjected to the median filter are inputted to the subtracter 41, and a difference value is taken.

This difference value is compared with a threshold value TH by a comparator 41+, and if the difference value is less than or equal to TH, a switch is made to output the output value a'' of the median filter lO, and if it is greater than TH, the output value a is not applied with the median filter. 12.

FIG. 4 shows the configuration of the low-pass filter 8 and the filter control section 9, and 4(1) is an example of the low-pass filter.

In Figure 4 (3), the low-pass filter 8 is multiplied by the multiplier 5.
2+"g, an adder 53, and a selector 54. Input pixels a, b, c, d, and e are multiplied by multipliers 521 to K.
respectively 1/2.1/8°1/8.1/8.1/
The values of 8 are multiplied together and added together by an adder 53, and the output value a' of the low-pass filter 8 is output.

The filter control section 9 is composed of 0R55, 0R56, and a delay device 57, and uses the 0R55 and the delay device 57 to determine whether the outputs of one block of the quantization section 5 are all O. If all zeros are present, it can be considered a non-coded block, otherwise it can be considered a coded block.

If the block is a coded block and the motion vector is not (0, O') in FIG. 3(2), the output of the AND 56 becomes 1, and this signal causes the selector 54 to otherwise, control is performed to select the unfiltered output a.

From the above, the low-pass filter 8 can be controlled.

〔Effect of the invention〕

As explained above, according to the present invention, in a motion compensated interframe coding method, in addition to motion vector information, insertion of a low-pass filter for local decoding output is controlled based on the conditions of non-encoding and coding blocks. , removes noise in the moving region without deterioration, and removes pulse-like noise in the static region by controlling the insertion of the median filter based on the error before and after inserting the median filter.

Image quality can be improved in both static areas.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing the basic configuration of the present invention. FIG. 2 is a diagram showing the configuration of an embodiment of the present invention. FIG. 3 is a diagram showing an example of the configuration of a median filter. FIG. 4 is a diagram showing an example of the configuration of a low-pass filter. FIG. 5 is a diagram showing a conventional technique. In the figure, ■ indicates that the frame memory 22 is a motion vector detection section;
3 is a variable delay unit, 4 is a calculation unit, 5 is a quantization unit, 6 is a decoding unit, 7 is a calculation unit, 8 is a low-pass filter, 9 is a filter control unit, 10 is a median filter, 11 is a filter control unit, 12 is a switch. Agent Patent Attorney Sada Igeta - (1'rt Q, C, cL, e i7) Medium heat blind (
1) Mehui 7nfilk Φ-dun I] Node〉7shihui J
Figure 1 shows the 4th generation of 20-bass fee W ri - 4

Claims (1)

[Claims] A frame memory (1) that stores locally decoded signals and generates image information of one frame before, and compares image information of the current frame with image information of the one frame before to calculate a motion vector. a motion vector detection section (2) for detecting a motion vector; a variable delay section (3) for generating a predicted value signal by applying a delay that varies according to the motion vector to the output of the first frame memory (1); A subtracter (4) that subtracts the predicted value signal from the image signal to generate a difference value signal, and a quantizer (5) that quantizes the difference value signal to generate a coded error signal. , a decoding unit (6) that decodes the encoded error signal, and an adder (7) that adds the decoded error signal and the previous prediction value signal to generate a locally decoded signal.
A motion compensated interframe coding system comprising: a low-pass filter (8) that removes high frequency components in the locally decoded signal; A filter control unit (9) that controls the characteristics of the low-pass filter (8) according to coding and non-coding conditions.
and a median filter (10) that removes pulse-like noise; a filter control unit (11) that controls insertion or non-insertion of the median filter (10) based on a difference value before and after the median filter (10); and a switch (12). ) is provided.