JPH08140095A - Coding device - Google Patents

Abstract

PURPOSE: To obtain a coded image whose quality is improved by performing an inter-frame prediction by an improved local decoded image. CONSTITUTION: An input frame 60 receives the processing for every pixel, the values for all the pixels (the present pixels) and 8 adjacent pixel values are obtained by a pixel retrieval 61, the values are delivered to a median calculation 62 via a line 66, the median value of these pixel values is determined, the value is delivered to a threshold comparison 63 via a line 67 and an inspection is performed for the value. When the absolute value of the difference of the present pixel and the median is less than the threshold of an adaptive filter, the present pixel value is replaced by the median value by the replacement 6 of the pixel by the median value. If the absolute value is other case except this, the present pixel value is not changed. When all the pixels within a frame receive this procedure, the frame for which a filtering is performed can be obtained.

Description

Detailed Description of the Invention

[0001]

FIELD OF THE INVENTION The present invention relates to the enhancement of coded video signals for digital storage and transmission media, and more particularly to the enhancement of images under video coding conditions with very low bit rates. It is a thing.

[0002]

2. Description of the Related Art Recently, various voluntary researches and standardization efforts have been developed due to the compression of the bandwidth of video signals. Various techniques have been explored for digitally encoding various forms of video signals using different bit rates.

There is spatial redundancy (ie, the pixels of an image within a frame have a value that is similar to the surrounding pixels) in a video sequence, and temporal redundancy (ie, for example, , If the same scene is shown in many frames of a video sequence,
The pixels of the image in one frame have the same characteristics as the pixels of the image in another frame). The compression is done by removing both spatial and temporal redundancy.

Currently established methods utilize both intra-frame and inter-frame coding to achieve extremely high compression ratios. In the case of intraframe coding, for example, discrete cosine transform (DCT) or Wa
A transform method, such as the velet transform, removes spatial redundancy when the pixels in the same image frame are processed. The results or coefficients are passed from the quantization process and then to variable length coding (VLC) to obtain compression information. The intra-frame method works only for image pixels located in the same frame of an image.

The compressed information can then be transmitted or stored. In order to obtain a decoded or reconstructed image at the receiver or decoder, the compressed information has to be passed to an inverse process, eg inverse VLC, inverse quantization, inverse DCT.

In interframe coding, pixel values present in another frame are used to predict or estimate pixels in one frame to remove temporal redundancy.
Forward prediction refers to prediction that uses pixels of a previous frame or preceding frame, and backward prediction refers to prediction that uses pixels of a future frame. Also, when a pixel is predicted from both the previous frame and the future frame, it is called bidirectional prediction.

In interframe coding, image pixels are predicted or estimated by utilizing pixels from other frames. The prediction does not always exactly match the original pixel, and for example, using DCT transform, quantization, and VLC, the difference or prediction that occurs in the same way as for intraframe coding. Final compression information is obtained by compressing or processing the error.

At the receiver, the decoding is done in the exact reverse process. As with intraframe decoding, the difference or prediction error is recovered through the reverse steps. Furthermore, the same decoded process result as the method utilized in the encoder is added with the recovered difference or error to obtain the final decoded or reconstructed image.

Motion-compensated (MC) prediction is also used for interframe coding, especially for moving picture sequences with moving scenes. Under these conditions, motion will occur, so by utilizing pixels from previous or future frames of the scene that have moved relative to their original position,
It allows for better matching and prediction of pixels within a frame. Motion vectors are used to indicate to the decoder where these scenes have moved and are sent along with compression information to indicate from which pixel region the interframe prediction should occur. Therefore, the decoder can accurately know where to perform motion compensation in order to obtain a decoded image.

[0010]

Very slow bit rates using current coding methods often result in insufficient bits available to code the entire image information, resulting in: Light and shade unevenness, corona noise (DCT
Due to distortion), or a patched state (due to Wavelet transform), will significantly damage the image information.

Current methods utilize prediction with locally coded images when interframe coding is used. Therefore, the result of the inter-frame prediction is further deteriorated due to the cumulative effect, since the result of the inter-frame prediction is predicted by using the locally encoded image which already contains a considerable amount of distortion.

It is an object of the present invention to provide a coding apparatus that obtains a coded image with improved quality by performing inter-frame prediction by using an improved locally decoded image.

[0013]

SUMMARY OF THE INVENTION With respect to the above-stated problems, the present invention introduces an adaptive filter in the inter-frame prediction loop to simplify the degradation or distortion that occurs in video coding with extremely slow bit rates. To provide a simple method.

At the time of encoding, the video signal is encoded as described in the prior art, ie the conversion process into a video signal (eg DCT or Wav).
elet transform), and further, after performing a quantization process, by performing variable length coding (VLC),
The compression information is obtained.

At the encoder, the locally decoded image is obtained by decoding the compressed information. It is this locally decoded image that is normally used for inter-frame prediction.

The invention proposes to adaptively filter these locally decoded images so that the distortions due to the coding are eliminated. Furthermore, subsequent inter-frame prediction is performed using the filtered or improved locally decoded image. This removes the cumulative distortion error in the inter-frame prediction loop.

Coding of the video signal is performed utilizing the prediction from the locally filtered image with improved filtering by providing an adaptive filter within the inter-frame prediction loop. As a result, a greatly improved coded image is directly obtained.

At the decoder, the decoding of the compressed information is
The process is exactly the same as for obtaining the locally decoded image described above. The obtained decoded image is also processed by the same filter to obtain a decoded image with improved quality. The decoded image that has been filtered is also
Used for subsequent interframe decoding.

[0019]

The best compression results are obtained by the combined use of intra-frame and inter-frame coding in moving picture sequences. First, it is determined whether the frame should be intra-frame encoded or inter-frame encoded. When the intra-frame coding is to be performed, first, the input video signal in the frame is processed by a transform process such as a discrete cosine transform, and then, quantization and variable length coding (VLC) are performed. ) Is applied and compressed information is obtained and then transmitted. In the encoder, the locally decoded image is obtained by performing an inverse process such as inverse VLC, inverse quantization, and inverse DCT on the compressed information. The locally decoded image is then passed through an adaptive filter, in particular to remove the coding distortion that occurs.

In the case of interframe coding, the input video signal is predicted using the previously decoded locally decoded frame. Next, as in the case of intra-frame coding, DCT, quantization, and VLC are performed on the difference or the prediction error as the prediction result to obtain the compressed information to be transmitted. In the encoder, the inverse process (eg inverse VLC, inverse quantization, inverse DC
By applying T), the recovered difference or prediction error is obtained. By adding these prediction errors to the same prediction process that was performed, an interframe locally decoded image is obtained. This locally decoded image is then filtered to remove distortion, and then
Stored for future predictions to be used.

At the decoder, the decoding is carried out in exactly the same process as for the decoding of the locally decoded image described above.

[0022]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below with reference to the drawings showing its embodiments.

FIG. 1 is a block diagram of an encoder using an adaptive filter. That is, the input sequence 1 is a series of image frames that form a video sequence.
The frames forming this sequence are processed frame by frame.

First, these frames are classified into I frames or P frames. I-frames and P-frames differ in how they are processed by the encoding algorithm. I-frames are coded using intra-frame coding techniques, i.e., regardless of adjacent frames. P-frames are coded using interframe coding techniques, that is, using motion estimation and compensation techniques to utilize information from neighboring frames and predict pixels to be coded. To be done. Most frames are coded using interframe coding techniques to reduce temporal redundancy. Intraframe coding is used periodically in a video sequence to support some other video playback features such as fast forward, fast rewind, random access, and to provide error robustness. The decision as to whether intraframe coding or predictive coding should be performed for a frame is made based on the frame number (or position) in the video signal sequence. The numbers (or positions) of I-frames and P-frames are predetermined. For example, if it is determined in advance that one I frame is provided for every 15 frames, the (15n + 1) th frame (where n = 0,
1, 2 ,. ． ． ), That is, the first, 16th, 32nd. ． ． Each frame becomes an I frame, and the rest are
It will be a P or B frame. The number of frames between I frames is called a picture group (GOP).
In the above example in which every (15n + 1) th frame is an I frame, the GOP length is 15.

To begin the encoding process, all input frames are sent via line 10 to the block sampling 2 routine. In Block Sampling 2, the frame is divided into 8 × 8 sized pixel blocks, which are then sent via line 12 to the DCT / Motion Compensation 3 routine. In this routine, if the pixel block to be encoded is from the I frame, no motion compensation is performed on the block, and the block uses the discrete cosine transform (DCT) to shift from the spatial domain to the frequency domain. To be converted. The pixel block to be encoded is
If from a P frame, the pixel block is subjected to a motion compensation process and then transformed from the spatial domain to the frequency domain by the Discrete Cosine Transform (DCT). The motion compensation process is performed by comparing the pixel block to be encoded with the locally decoded frame from the locally decoded frame memory 7. The transformed coefficients are then quantized 4 via line 13 for quantization.
And the quantized coefficient is obtained. These quantized coefficients are then sent via line 14 to the run length variable length coding 5 routine. In this routine, a run-length scan is performed and then Huffm
Applying variable length coding using the an coding technique produces output coded information 22, which is then sent to a decoder or stored on a storage medium.
The quantized coefficients from quantization 4 are also sent via line 20 to the dequantization / inverse DCT 6 routine and also to the inverse motion compensation 8 routine to produce the locally decoded frame. It will be. This locally decoded frame is further sent via line 16 to the routine of the adaptive filter 17 to undergo the adaptive filtering process and then stored in the locally decoded frame memory 7 via line 21. The adaptive filtering process of adaptive filter 17 is further detailed in the following paragraphs related to FIG.

FIG. 2 is a flow chart for the decoder of this embodiment. That is, the encoded information is line 4
It is input via 1. Next, quantized DCT coefficients are obtained by subjecting the encoded information to variable length decoding / run length decoding 36. In the inverse quantization 37, the inverse quantization of the quantized DCT coefficient is performed, and the coefficient is sent to the motion compensation / inverse DCT 38 via the line 43. P
Performing motion compensation on the frame and then performing inverse DCT on the coefficients yields a reconstructed pixel block, which is used to form the reconstructed frame. To be done. For I-frames, the reconstructed pixel block is obtained by performing only the inverse DCT on the coefficients. The reconstructed frame is sent to the adaptive filter 39 via the line 44, and the coding noise generated in the reconstructed frame is removed to obtain the reconstructed sequence that is finally output. . The filtered frame is also sent to local decoding frame memory 40 via line 45. The filtered frame of the local decoding frame memory 40 is used in the motion compensation / inverse DCT 38 motion compensation process. The adaptive filter 39 is the same as that utilized in the encoder of FIG. 1 and will be further detailed in the next paragraph related to FIG.

FIG. 3 is a flowchart of adaptive filtering using the median filter of this embodiment.
That is, the input frame 60 is processed pixel by pixel. In the pixel search 61 routine, for every pixel (current pixel), its value and the values of the eight adjacent pixels are obtained, and through line 66,
It is sent to the median calculation 62 routine. The median calculator 62 routine determines the median value of these pixel values and sends the threshold comparison 6 via line 67.
It is sent to 3 and is inspected. If the absolute value of the difference between the current pixel and the median is less than the adaptive filter threshold, the replace pixel to median value 64 routine replaces the current pixel value with the median value, and otherwise. , The current pixel value is unchanged. Therefore, when all the pixels in the frame have undergone the above procedure, the processed (filtered) frame is obtained.

[0028]

INDUSTRIAL APPLICABILITY Introducing an adaptive filter in the prediction loop, in particular for reducing coding distortion, enables subsequent frame prediction using an improved locally decoded image. This has a cumulative effect in reducing the distortion generated in the prediction frame, and thus greatly contributes to the improvement of the image quality. The invention has the effect of directly improving the quality of the coded image, especially in techniques where interframe coding is used. INDUSTRIAL APPLICABILITY The present invention can also be used in future image coding standards with low bit transmission rates.

[Brief description of drawings]

FIG. 1 is a block diagram of an encoder using an adaptive filter.

FIG. 2 is a flowchart relating to the decoder of this embodiment.

FIG. 3 is a flowchart of adaptive filtering using the median filter of this embodiment.

[Explanation of symbols]

 7 Local Decoding Frame Memory 17 Adaptive Filter 25 Decoder 39 Adaptive Filter 40 Local Decoding Frame Memory

Claims (6)

[Claims] 1. Means for receiving an input video signal; means for encoding a video signal into compressed information;
Means for decoding the compressed information into a locally decoded video signal, means for subjecting the locally decoded video signal to a filtering process, and utilizing the result of the filtered locally decoded video signal. And a means for performing inter-frame prediction. 2. The filtered locally-decoded video signal of claim 1, wherein the video signal encoding means is means for subjecting the video signal to an interframe prediction process. Means for calculating a difference or prediction error between the inter-frame prediction process result and the video signal, the difference or prediction error being subjected to a compression process to obtain compression information. The encoding device according to claim 1, further comprising: 3. Means for receiving compressed information, means for decoding compressed information into a decoded video signal, means for subjecting said decoded video signal to a filtering process, and filtering. And a means for performing inter-frame prediction by using the result of the decoded video signal. 4. A means for decoding compressed information, for decompressing the compressed information composed of an inter-frame predictive coded video signal to obtain a difference or a prediction error,
A means for adding the difference or prediction error to the result of an inter-frame prediction process performed using the result of the filtered decoded video signal according to claim 3. The encoding device according to Item 3. 5. A filtering process means for providing an evaluation result based on a video signal value around each video signal; a means for checking a difference between the evaluation result and the video signal value; Means for making a comparison with a predetermined threshold value based on the difference; means for determining whether the video signal needs to undergo a filtering process based on the comparison; and the video signal. The encoding device according to claim 1 or 3, further comprising means for receiving a filtering process based on the comparison result. 6. The encoding according to claim 1, wherein the video signal is composed of pixels from a frame of pixels, and the frame is composed of a field of pixels. apparatus.