KR100272818B1 - Image datd encoding device - Google Patents

Abstract

According to the present invention, in the moving picture data decoding apparatus, even if there is a transmission error in which the coding error cannot be corrected on the decoding side, the missing image can be corrected inconspicuously.

When a code error that cannot be corrected is detected, it is determined whether the motion vector of the motion vector corresponding to the moving picture data outputted with the error detection signal is larger than a predetermined value, and when the motion amount is large, image data and predictive image data are synthesized. Irrespective of the magnitude of the movement of the decoded moving image data S21, the deletion due to a sign error can be corrected inconspicuously.

Description

Video data decoding device

1 is a block diagram showing the construction of a first embodiment of a moving picture data encoding apparatus of the present invention.

2 is a schematic diagram that provides a description of interframe motion prediction.

3 is a block diagram showing the construction of a first embodiment of a moving picture data decoding apparatus of the present invention.

4 is a schematic diagram providing a description of block interpolation.

5 is a schematic diagram providing a description of the in-frame interpolation process.

6 is a schematic diagram provided in the description of another embodiment.

* Explanation of symbols for main parts of the drawings

DESCRIPTION OF SYMBOLS 1: Video data coding apparatus 20: Video data decoding apparatus

21: input terminal 22: buffer circuit

23: sign error detection / correction circuit 24: demultiplexer circuit

25 variable length decoding circuit 26 inverse quantization circuit

27: discrete cosine inverse conversion circuit 28: frame data generation circuit

29: frame memory 30,31,33: switching circuit

32: motion vector memory 34: motion compensation circuit

40: interpolation circuit 41, 42: delay circuit

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving picture data decoding apparatus, and is suitable for the application of moving picture data from a recording medium such as a compact disc, a so-called digital audio taperecorder (DAT) cassette, or a hard disc.

Conventionally, since a large amount of information is required to digitally record a moving picture, a recording medium having a very high continuous transfer rate is required for recording / reproducing it.

For example, when digitally recording a video signal by NTSC, recording / reproducing is performed using a so-called video disc having a large amount of recording information.

However, in order to record moving image data of the generated information amount as in the case of a video disc for a long time on a smaller (ie less recording information amount) recording medium, the video signal is encoded with high efficiency and the read signal is efficiently Means for decryption are indispensable.

In order to meet such demands, a high efficiency coding method of an image signal has been proposed, and one of them is a moving picture expert group (MPEG) method.

The MPEG method first takes a difference between images in order to reduce redundancy in the time axis direction, and then performs discrete cosine transform (DCT) in order to reduce the redundancy in the spatial axis direction.

By the way, in such an MPEG system, a decoding device is provided at the front of the decoding. When an error that cannot be corrected by the error code correction capability of the decoding device occurs, a partial missing of the image occurs at that point, and the image is displayed as it is. If you do this, you'll be in trouble.

Therefore, in order to correct partial missing of this image, two conventional methods have been proposed.

One is a partial image positioned at the same position as the missing image portion, and is an interframe interpolation method of correcting by inserting a past frame image into the missing portion, and the other is an image interpolated from surrounding pixels of the missing image portion. In-frame interpolation that modifies by embedding in the missing part.

By the way, the interframe interpolation method is effective in an image part with little movement, but there is a problem in that the missing part and the peripheral part cannot be smoothly connected in the part where the motion is severe, and the result is not clear, and a good result is not necessarily obtained.

In addition, the intraframe interpolation method can correct the lack of notice in the part of the image that has a large movement, as opposed to the interframe interpolation method. There was.

The present invention has been made in view of the above, and even when a decoding error occurs in which a coding error cannot be corrected on the decoding side, the image can be decoded inconspicuous regardless of the magnitude of the missing image. It is an object of the present invention to propose a moving picture data decoding apparatus.

In order to solve such a problem, according to the present invention, in a moving picture data decoding apparatus which decodes sequentially inputted moving picture data (S21), a code error of the moving picture data (S21) is corrected and the reproduced digital signal (S22) is output. At the same time, error detection / correction means 23 for outputting an error detection signal S23 and a reproduction digital signal S22 are outputted to the motion vector data S30 when a code error that cannot be error-corrected is detected in the moving image data S21. ) And the demultiplexing means 24 separating the image data S24 and the error detection signal S23 and the motion vector data S30, and the error detecting / correcting means 23 can correct the code error. In the case where the predictive image data S33 is generated based on the motion vector data S30, and the error detection / correction means 23 cannot correct the code error, the corresponding picture error is detected in the moving picture data S21. An error based on predicted image generating means 31, 32, 33, 34 for generating predictive image data S33 based on the corresponding motion vector S31, and an error detection signal S23 and an image S24. Interpolation image generating means 40 for generating interpolation image data S43 corresponding to moving image data S21 in which an uncorrectable code error is detected, and error detection / correction means 23 can correct the code error. If there is a code error that synthesizes the image data S24 and the predictive image data S33 and detects a code error that cannot be error corrected, the motion vector S41 determined corresponding to the moving image data S21 to which the error detection signal is output. And image synthesizing means 28 and 29 for synthesizing with either the image data S24, the predictive image data S33, or the interpolation image data S43 based on the amount of motion.

If a code error that cannot be error corrected is detected when decoding the moving picture data S21, the image is based on the motion amount of the motion vector S41 corresponding to the moving picture data S21 to which the error detection signal S23 is output. By combining the interpolated image data S43 or the predicted image data S33 with the data S24 and outputting them, the missing due to a sign error can be inconspicuously corrected regardless of the movement of the decoded moving image data S21. .

EXAMPLE

Hereinafter, a first embodiment of the present invention will be described with reference to the drawings.

In FIG. 1, 1 denotes a moving image data encoding apparatus (encoder) as a whole, and input image data S1 obtained by converting an analog moving image signal into digital data is input from the input terminal 2. FIG.

At this time, the input image data S1 is composed of an intra frame I, a predirect frame P, and a bidirectional frame B as shown in FIG.

Here, the intra frames I ₀ and I ₁ are frames that are data-compressed and transmitted only within the frame, and the predictive frames P ₀ , P ₁ ... Are frames that are motion predicted from one direction, and the bidirectional frame B (B ₀ , B ₁ , B ₂ , B ₃ ...) Are frames whose motion is predicted from both directions.

The difference data generation circuit 3 inputs the input image data S1 from the input terminal 2 and simultaneously receives the front frame image data S2 of the front frame stored in the frame memory 4 from the frame memory 4. It is made up of input.

Here, the difference data generation circuit 3 obtains the difference between the input image data S1 and the front frame image data S2 to generate the difference data S3, and to the discrete cosine transform DCT circuit 5. Output

The discrete cosine transform circuit 5 uses the two-dimensional correlation of the image and discrete cosine transforms the difference data S3 in units of small blocks, and converts the resulting transform data S4 into the quantization circuit Q (quantizes) 6. To print.

When the quantization circuit 6 quantizes the transform data S4 to a predetermined quantization step size, the quantization circuit 6 outputs the quantization data S5 obtained at the output terminal to the variable length coding (VLC) circuit 7. .

The variable length encoding circuit 7 performs variable length encoding processing on the quantized data S5 and outputs the variable length encoded data S6 to the multiplexer 8.

When the multiplexer 8 multiplexes the motion vector data S7 input from the encoder 9 to the variable length coded data S6, the multiplexer 8 transmits the transmitted data S8 via the buffer circuit 10.

The moving picture data encoding apparatus 1 also has a local decoding circuit 11, which locally decodes the quantized data S5 transmitted as the transmission data S8 and supplies it to the frame memory 4.

When the local decoding circuit system 11 inputs the quantization data S5 to the inverse quantization circuit Q- ¹ , 12, the local decoding circuit system 11 dequantizes the quantization data S5 to a representative value, converts the inverse quantization data S10, and then converts the quantization data S5 to the inverse quantization data S10. The transformed data is decoded and supplied to the discrete cosine inverse transform (DCT- ¹ ) circuit 13.

The discrete cosine inverse conversion circuit 13 converts the inverse quantization data S10 decoded by the inverse quantization circuit 12 into the decoded image data S11 by a conversion process with the discrete cosine conversion circuit 5, and the frame data generation circuit ( 14).

Here, the frame data generation circuit 14 adds the frame image data S2 and the decoded image data S11 fed back from the frame memory 4 to restore the image data output as the transfer data S8, and restores the frame memory. It is stored in (4) sequentially.

In addition, the moving picture data encoding apparatus 1 inputs the input image data S1 to the motion vector calculating circuit 18 to obtain a motion vector. The moving picture data encoding apparatus 1 transmits the input image data S1 to the motion compensation circuit 19 and the encoder 9 as the motion data S15. It is to supply.

Here, the motion compensation circuit 19 reads the decoded image data S16 from the frame memory 4, and simultaneously performs motion prediction data S17 for motion compensation on the decoded image data S16. It is supposed to output to.

In addition, the encoder 9 encodes the motion data S15 obtained by the motion vector calculating circuit 18 and outputs it to the multiplexer 8 as the motion vector data S7.

20 shows a moving picture data decoding apparatus (decoder) as a whole, and reproduces the read data S21 read from the recording medium from the input terminal 21 to the code error detection / correction circuit. It is supposed to be input in (23).

Here, the code error detection / correction circuit 23 detects and corrects an error included in the image data from the reproduction data S21 and outputs the corrected reproduction image data S22 to the demultiplexer circuit 24.

In addition, the code error detection / correction circuit 23 outputs a switching signal S23 for controlling the switching of the output image data when detecting a block that cannot correct the code error.

The demultiplexer circuit 24 separates the motion vector data from the reproduced picture data S22 and supplies it to the variable length decoding circuit VLC- ¹ 25 as the differential picture information data S24, and in the variable length coding circuit 7 The decoded image data S25 before encoding is decoded and supplied to the dequantization circuit Q ^-1 26.

The inverse quantization circuit 26 inversely quantizes the decoded image data S25 to a representative value and converts it into inverse quantization data S26. It converts into decoded image data S27 and outputs it to the frame data generation circuit 28.

In addition, the frame data generation circuit 28 adds the decoded data S27 to the motion compensation data S28 read out from the frame memory 29 to decode the decoded image data S29 and passes through the switching circuit 30. It outputs from the frame memory 29.

Here, the switching circuit 30 is on / off controlled by the switching signal 23 supplied from the code error detection correction circuit 23 and cannot be corrected by the reproduction data S21 in the code error detection correction circuit 23. When a sign error is detected, the decoded image data S29 is not stored in the frame memory 29.

When the demultiplexer circuit 24 separates the vector data from the reproduced image data 22, it is supplied to the motion vector memory 32 via the switching circuit 31 as the current motion vector data S30.

Here, the switching circuit 31 holds the motion vector for the past three frames when the code error detection / correction circuit 23 detects no code error that cannot be corrected by the reproduction data S21. At 32, the current motion vector data S30 is supplied.

The switching circuit 33 is controlled to be switched by the switching signal S23 supplied from the code error detection / correction circuit 23, and a code error that cannot be corrected is not detected in the reproduction data S21 of the current frame. Next, the current motion vector S30 is supplied to the motion compensation circuit 34, or, if a code error that cannot be corrected is detected, the previous motion vector data S31 held in the motion vector memory 32 is moved. It is supplied to 34.

When the motion compensation circuit 34 inputs the reference frame data S32 from the frame memory 29, the motion compensation circuit 34 generates a predictive block image based on the current or previous motion vectors S30 and S31, and generates the frame memory as the predictive image data S33. It is stored in (29).

When an uncorrectable code error occurs, the frame memory 29 replaces and stores the predictive picture data S33 generated by the previous motion vector S31 as the picture data of the corresponding block portion. It outputs as output image data S34.

When the amount of motion of the previous motion vector S31 is large with respect to the predetermined value, the frame memory 29 replaces and stores the interpolation image data S43 newly input from the interpolation circuit 40 with the predictive image data S333. Then, it outputs as the output image data 34.

The interpolation circuit 40 inputs the motion vector S41 from the motion vector memory 32 via the delay circuit 41 and inputs the switching signal S23 via the delay circuit 42 and the motion vector data S41. If the amount of motion is large with respect to the preset threshold, the pixel A of the adjacent block when X is used as the image data S42 (the missing pixel in FIG. 4) for 4 pixels read from the frame memory 29. Interpolation image data S43 is generated on the basis of (B), (C), (D).

The interpolation circuit 40 requires four pixels of image data (a, b, c, d) located in four blocks at the top, bottom, left, and right sides of the block R of interest in order to perform the interframe processing of the corresponding frame. The decoding of the block adjacent to the lower side waits for the end and interpolates the pixel data.

For reference, the interpolation circuit 40 starts the intra-frame interpolation process soon after the inter-frame interpolation process by the motion compensation circuit 34 is completed, when the block of interest R is located at the lowermost column of the screen.

Here, the delay circuits 41 and 42 may include the motion vector data S41 and the interpolation circuit 40 so as to correct the past block of one example (one slice) of the block in which the current decoding is finished or the inter-frame modification is finished. The switching signal S23 is delayed.

In the above configuration, the moving picture data decoding apparatus 20 reads the reproduction data S21 sequentially from the recording medium and inputs it to the code error detection and correction circuit 23 via the buffer circuit 22.

If the code error detection correction circuit 23 can correct the code error of the reproduction data S21, the moving picture data decoding apparatus 20 corrects the corrected image data S22 corrected based on the error detection correction code. The demultiplexer circuit 24, the variable length code decoding circuit 25, the inverse quantization circuit 26, and the discrete cosine inverse conversion circuit 27 are sequentially supplied to the frame data generation circuit 28.

The frame data generation circuit 28 adds the decoded data S27 to the motion compensation data S28 obtained by motion compensation of the intra frame 10 stored as a reference frame in the frame memory 29 to decoded image data S29. ) Is supplied to the frame memory 29 via the switching circuit 30 to sequentially decode the image data of the predictive frame P0 and the bidirectional frame B0, B1... As a decoded image data S344. Output

On the other hand, the moving picture data decoding apparatus 20 transmits the switching signal S23 to the switching circuit 30 when the code error detection correction circuit 23 detects a block that cannot correct the code error of the reproduction data S21. 31) and write-out of the decoded image data S29 and the current motion vector S30 into the frame memory 29 and the motion vector memory 32 is stopped.

Further, the moving picture data decoding apparatus 20 outputs the switching signal S23 to the motion vector memory 32 and the switching circuit 33 and moves in the corresponding block of the past frame stored in the motion vector memory 32. The vector is read out, supplied to the motion compensation circuit 34, and replaced with the corresponding pixel by the predictive image data S23 predicted from the past motion vector.

At this time, the interpolation circuit 40 determines whether the motion amount of the motion vector data S41 input through the delay circuit 41 is large with respect to the predetermined value, and if the motion amount is small with respect to the predetermined value, the frame memory 29 The interpolated interpolated frame image is output as the output image data S34 by replacing the pixel data x of the missing pixel X with the predictive image data S33.

On the other hand, the interpolation circuit 40 has four pixels A, B, of the block adjacent to the block of interest R, in which the pixel X in which a sign error occurs that cannot be error corrected when the amount of motion is large for a predetermined value. Image data of the pixel X of interest is generated from C and D, and output as interpolation image data S43.

Here, the four pixels A, B, C, and D in the adjacent block are pixels located in the same row or in the same line as the pixel X of interest and adjacent to the pixel X of interest, and each pixel value is a, b, c, and d.

If the distance between the four pixels A, B, C, and D and the pixel K of interest is LA, LB, LC, and LD, the pixel value of the pixel X of interest is represented by the following equation.

Is obtained by linear interpolation and is output to the frame memory 29 as interpolation image data S43.

At this time, the memory 29 interpolates the interpolated image generated by interpolation in the frame instead of the predicted image data S33 in which the pixel value x of the pixel of interest X in which a code error which cannot be error corrected has occurred is stored once. The data S43 is newly rewritten and output as the output image data S34.

On the other hand, when the block of interest R containing the pixel of interest X having a protection error that cannot be error-corrected is the outermost block of the screen, or one of the adjacent blocks has a code error that cannot be error-corrected. In the case of a block, since the four pixels A, B, C, and D cannot be referred to from the adjacent block, the pixel value x is obtained by using pixels of the opposite neighboring block for the adjacent block that cannot be detected.

For example, in the case where the pixel value b of the pixel B cannot be obtained because the adjacent block on the left side cannot be referred to in FIG.

Substituting into (1), the pixel value x of the pixel X of interest is obtained.

In addition, when it is impossible to refer to adjacent blocks located at both the left and right neighbors of the block of interest R including the pixel of interest X having an error code that cannot be error corrected.

The pixel value x is obtained, and the image of the missing pixel is corrected inconspicuously with respect to the peripheral portion.

According to the above configuration, in the moving picture data decoding apparatus which decodes and reproduces the moving picture data in units of blocks, when a code error that cannot be error corrected is detected in the reproduced moving picture data, the previous frame is based on the motion vector of the previous frame. If the image of the corresponding block is once replaced in the predicted image generated from the image, and the motion vector of the motion vector used for the prediction is large, it is replaced by the interpolated image interpolated from the surrounding pixels, regardless of the magnitude of the motion of the missing image. The missing image can be corrected so that the missing due to the sign error is not conspicuous.

In addition, in the above-described embodiment, the case in which the motion is predicted in units of frames is described. However, the present invention is not limited thereto, and the present invention can also be applied to a system in which motion is predicted in units of fields.

In addition, in the above-described embodiment, a case of applying to a system using discrete cosine transform (DCT) and inter-frame motion prediction has been described. However, the present invention is not limited to this, and a moving picture which performs inter-frame prediction processing is described. It is suitable for application to a data decoding apparatus.

In addition, in the above-described embodiment, when frame interpolation of image data in which a code error that cannot be error corrected occurs is interpolated by image data which is motion compensated by a motion vector of the past corresponding to the image data, a description will be given. Although the present invention is not limited to this, the present invention may be replaced with pixel data in the corresponding block of the past frame.

In addition, in the above-described embodiment, the case where inter-frame or inter-frame interpolation processing is performed by using the motion vector of the past frame when a sign error that cannot be error-corrected has been described, the present invention is not limited to this. May be interpolated using the motion vectors of adjacent blocks in the frame.

For example, for the block of interest R, the motion vector of the block on the left side or the upper side where decoding has been completed may be used.

In addition, in the above-described embodiment, the equation (3) can not be referred to when adjacent blocks located in both neighbors of the block of interest R including the pixel of interest X that cannot be error-corrected can be referred to. The case where the pixel value x of the pixel X of interest is obtained is described. However, the present invention is not limited to this, and in the case where adjacent blocks adjacent to each other in the vertical direction of the block R of interest cannot be referenced, The pixel value x of the pixel X of interest can be obtained.

Incidentally, in the above embodiment, the sequence of the configuration shown in FIG. 2, that is, I ₀ , B ₀ , B ₁ , P ₀ , B ₂ , B ₃ , P ₁ , B ₄ , B ₅ , I ₁ . For example, the present invention is not limited thereto, and the present invention can be applied to sequences in various combinations of the intra frame 1, the predirect frame P, and the bidirectional frame B. .

In the above embodiment, a case has been described in which the motion vector memory 32 stores motion vectors for the past three frames. The present invention is not limited to this and can be widely applied to the case of storing motion vectors of a plurality of frames in the past.

As described above, according to the present invention, when a code error with an error correction is detected, the image data and the predictive image data or the image data are synthesized in accordance with the motion amount of the motion vector corresponding to the moving image data in which the code error is detected. By decoding, the missing portion of the image due to a sign error can be smoothly corrected with respect to the surrounding image, regardless of the magnitude of the movement of the decoded image.

Claims (1)

A moving picture data decoding apparatus which decodes sequentially inputted moving picture data, corrects a sign error of the moving picture data, outputs a reproduction digital signal, and detects a sign error that cannot be error corrected in the moving picture data. Error detection / correction means for outputting a signal, demultiplexing means for separating the reproduced digital signal into motion vector data and image data, input the error detection signal and the motion vector data, and the error detection / correction means When the code error can be corrected, predictive image data is generated based on the motion vector data, and when the error detection / correction means cannot correct the code error, the corresponding code error is detected in response to the moving picture data. Generate predictive image data based on the determined motion vector Is a predictive image generating means, interpolated image generating means for generating interpolated image data corresponding to moving image data in which the code error which is not error corrected based on the error detection signal and the image data is detected, and the error detection / correction. The means synthesizes the image data and the predictive image data when the means can correct a code error, and when a code error that cannot be error corrected is detected, an error detection signal is determined to correspond to the moving picture data outputted. And image synthesizing means for synthesizing either the image data, the predictive image data, or the interpolation image data based on the amount of motion.