JPH09238351A - Image coding device - Google Patents

Abstract

(57) [Abstract] [PROBLEMS] To reduce deterioration of decoded image quality that occurs when a special image is input, without significantly increasing the cost.
Improve image quality. SOLUTION: A first image memory 16 for accumulating a forward reference image, a first motion vector detecting means 19 for reading out image data of the first image memory and detecting a motion vector, and Second for accumulating the reference image of the direction
Image memory 17, a third image memory 18 for accumulating an extended forward reference image, image data of the second image memory or the third image memory, and detects a motion vector. In the case of the forward prediction, the second motion vector detection means refers to the image data of the third image memory to obtain the forward reference image from the reference image expanded in the forward direction. The motion vector is detected.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an improvement of an image coding apparatus which compresses image data for transmission and recording of digital image signals.

[0002]

2. Description of the Related Art Several systems have been put into practical use as a system used for encoding (compressing) a huge amount of data of a moving image, but as a representative system for performing bidirectional predictive encoding. , MPEG2 (Moving Picture Expert Gr
oup-2) is available. This MPEG2 is a DCT (Discrete Co
This is a method that combines data compression technologies such as orthogonal compression coding technology using sine transform, bidirectional motion compensation predictive coding technology, and variable length coding technology.

The MPEG2 coding system will be described below.

FIG. 3 is a block diagram showing the basic structure for realizing the MPEG2 coding system. In addition, MPEG
For the explanation of No. 2, "Nakashima et al.," Video Compression ", TV Society Journal, Vol.49, No.4 pp.435-466".
Is quoted.

The digital image signal input from the input terminal 1 is an I picture for intra-frame coding in the screen rearrangement circuit 2, a P picture which is a forward prediction coding prediction screen, and a bidirectional prediction coding prediction screen. The prediction reference screen is rearranged so that it is encoded first in screen units such as a certain B picture.

In the intra-frame coding mode, these image data are encoded by a macro block (for example,
The data in 16 pixels × 16 lines) is sent to the DCT circuit 4 via the first mode selection switch 10 in DCT block (for example, 8 pixels × 8 lines) unit, and is DCT-converted as it is.

On the other hand, in the motion compensation predictive coding mode, the input image data to be coded and the image memory 31 are input.
A motion vector is detected by the motion detection circuit 32 from the reference image stored in, and a prediction image of the coding target image is created by the motion compensation prediction circuit 30 based on this motion vector. The difference between the image data to be encoded is calculated by the difference circuit 3 to obtain a prediction error signal, and the DCT circuit 4 performs DCT conversion on this prediction error signal.

The DCT transform coefficient is quantized by the quantizing circuit 5 on the basis of the quantizing coefficient determined by the rate control circuit 9 for controlling by monitoring the amount of data in the buffer memory 7 which will be described later, It is input to the variable length coding circuit 6 together with the motion vector and coding mode information. The variable length coding circuit 6 allocates a code having a high transmission efficiency by utilizing the statistical property of the input data and performs data compression. The variable-length coded data is stored in the buffer memory 7 and output in the MPEG format.

Further, in the case of I picture and P picture, since it is necessary to use as a reference screen for motion compensation prediction, the quantized information is the inverse quantization circuit 11 and inverse DCT.
The circuit 12 restores the data before the input to the DCT circuit 4. Further, in the case of the P picture, the predicted image data which is the output data of the motion compensation prediction circuit 30 via the second mode selection switch 15 is added by the addition circuit 13 to perform local decoding, and the image memory 31. Stored in.

In the coding of each screen, various coding modes are selected from the generated information amount in the mode selection circuit 33 in units of macroblocks (blocks in which a plurality of the DCT blocks are collected).
Determined by The type of mode differs depending on each picture, but the decision method is to compare the root-mean-square errors in the macroblocks for each mode and select the smallest mode.

In the case of the B picture, since it is bidirectional prediction, a circuit for forward prediction and a circuit for backward prediction are required, and different image memories are referred to. Therefore, only P picture, that is, forward prediction is required. It requires twice the circuit scale as compared with the case.

[0012]

In general, bidirectional predictive coding is performed to reduce the prediction error to improve the coding efficiency. However, video with a lot of motion such as sports broadcasting and fast panning of the camera are frequently performed. In video and the like, improvement in coding efficiency due to bidirectional predictive coding cannot be expected so much, and deterioration of decoded image quality cannot be avoided.

In order to suppress this deterioration, it is necessary to reduce the prediction error even for fast-moving images.
For that purpose, it is effective to expand the search range when performing motion compensation prediction and increase the accuracy of the motion vector in order to improve the coding efficiency.

However, in detecting the motion vector,
It seems that, like the block matching method used in the MPEG2 standardization process, a method of searching for an image having the smallest cumulative sum of the differences from the input image among the reference images in the search range is generally used. When the search range is expanded by the method, a motion vector detection circuit (motion detection circuit 3 in FIG.
There was a problem that the amount of hardware in 2) was significantly increased, resulting in a high cost device.

(Object of the Invention) An object of the present invention is to improve the image quality by reducing the deterioration of the decoded image quality that occurs when a special image is input, without significantly increasing the cost. An object is to provide an image encoding device.

[0016]

In order to achieve the above object, the present invention provides a first image memory for storing a forward reference image, and image data in the first image memory, A first motion vector detecting means for detecting a motion vector, a second image memory for storing a backward reference image, and a third image memory for storing an expanded forward reference image.
Image memory and second motion vector detecting means for detecting the motion vector by reading the image data of the second image memory or the third image memory, and in the case of bidirectional prediction, the first motion The vector detecting means detects the motion vector from the forward reference image by referring to the image data of the first image memory, and the second motion vector detecting means refers to the image data of the second image memory. By doing so, the motion vector from the backward reference image is detected, and in the case of forward prediction, the first motion vector detection means operates in the same manner as in bidirectional prediction, and the second motion vector detection means. Are configured to detect a motion vector from a reference image expanded in the forward direction by referring to the image data of the third image memory.

That is, a third image memory for accumulating the expanded forward reference image is provided, and in the forward prediction, the second motion vector detecting means used at the backward prediction is used as the third motion vector detecting means. It is used together with the image memory of to expand the search range.

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, the present invention will be described in detail based on illustrated embodiments.

FIG. 1 is a block diagram showing the arrangement of an image coding apparatus according to the first embodiment of the present invention.

In FIG. 1, 1 is an input terminal for inputting a digital image signal, 2 is an image rearrangement circuit for rearranging the input images on a screen-by-screen basis so that the prediction reference screen is encoded first, and 3 is the image. A difference circuit 4 for obtaining an error signal between the output signal of the rearrangement circuit 2 and the prediction screen output from the motion compensation prediction circuit 14 which will be described later is an error signal obtained by the difference circuit 3 or the image rearrangement circuit 2 DCT circuit for transforming the output image data of the image into the spatial frequency domain, 5
Is a quantizing circuit for quantizing the coefficient after conversion by the DCT circuit 4, and 6 is a variable length coding circuit for allocating a code with good transmission efficiency by utilizing the statistical property of the quantized data. Is. Reference numeral 7 is a buffer memory for accumulating data so that the output data at the bit stream output terminal 8 has a constant rate and monitoring the generated code amount. Here, the code amount is sent to the rate control circuit 9 to perform the rate control. The circuit 9 controls the amount of generated code by controlling the quantization coefficient of the quantization circuit 5.

Further, 11 and 12 are the DCT circuits 4,
Inverse quantization circuit and inverse D, which is an inverse conversion circuit for returning the coded data (image data) quantized by the quantization circuit 5 to the state of the input data of the DCT circuit 4.
It is a CT circuit.

When the switch 15 is in the motion compensation mode, the output of the inverse DCT circuit 12 is combined with the prediction screen which is the output of the motion compensation prediction circuit 14 in the combination circuit 13,
The original image is decoded, sent to the first to third image memories 16 to 18, and stored as a reference screen.

Such a main encoding system has the same configuration as the system shown in FIG. 3 as a conventional example.

Hereinafter, a method of forming a motion compensation prediction screen and a constituent circuit which are cores of the embodiment of the present invention will be described in detail.

An image serving as a reference screen is stored in an image memory (16 to 18) according to a position with respect to an image to be encoded.
Is different. FIG. 2 shows the range of image data stored in this image memory.

In the case of a B picture for bidirectional prediction, the reference screen for forward prediction is centered on the data of the encoding target image macroblock (range shown in 41 of FIG. 2: 16 pixels × 16 lines), and The past I picture or P picture image data in a range wider than the block (range indicated by 42 in FIG. 2: 32 pixels × 32 lines) is reproduced and stored in the first image memory 16 as a reference screen.

In the first motion vector detection circuit 19, the image data of the reference screen stored in the first image memory 16 and the image data of the image macroblock to be encoded are compared to detect the motion vector MV1. There is. As the detection method of the motion vector MV1, the block matching method described above is used. This motion vector MV1 is sent to the motion compensation prediction circuit 14.

On the other hand, the future reference picture for backward prediction is stored in another second image memory 17 with the same size as in forward prediction. Then, similarly to the case of the forward prediction, the second motion vector detection circuit 20 obtains the motion vector MV2, and the motion vector MV2 is sent to the motion compensation prediction circuit 14.

The motion compensation prediction circuit 14 creates a prediction motion compensation prediction screen from the past reproduced image and motion vector MV1 and the future reproduced image and motion vector MV2.

Regarding which data of the forward direction / rearward direction / bidirectional direction is used to create the prediction screen, the square prediction error is obtained for all the prediction modes, and the minimum mode is selected by the mode selection circuit 21. The information is used as the motion compensation prediction circuit 1
4

Further, in the first and second motion vector detection circuits 19 and 20, in order to determine the coding mode of intra-frame or motion-compensated prediction, the mean square deviation of the image macroblock data to be coded and the motion prediction. The mean square error of time is obtained and sent to the mode selection circuit 21. The mode selection circuit 21 selects an optimal encoding method from these pieces of information and switches between the first mode selection switch 10 and the second mode selection switch 15. The two mode selection switches are interlocked, In in FIG. 1 indicates the intra-frame coding mode, and MC indicates the motion compensation predictive coding mode.

In the case of a P picture for forward prediction, the first
Similar to the case of forward prediction of the B picture, the past reproduced image is stored in the image memory 16 of No. 1, and the image of the reference screen stored in the first image memory 16 is stored in the first motion vector detection circuit 19. The data and the image data of the encoding target image macroblock are compared, and the motion vector MV1
Is detected. Further, in the third image memory 18, a wider range in the screen stored at the same time as the reference screen stored in the first image memory 16 (range 43 in FIG. 2: 48 pixels × 32 lines excluding the central portion) ) Image data is stored. Second
The motion vector detection circuit 20 of the third image memory 1
The reference screen stored in 8 and the image data of the encoding target image macroblock are compared, and the motion vector MV2 is detected. Then, these motion vectors MV1 and M
V2 is supplied to the motion compensation prediction circuit 14.

Here, in order to obtain the optimum motion vector and the optimum reference screen, the first motion vector detecting circuit 19 determines which data is used to create the motion compensation prediction image.
Then, the square prediction error is obtained in the second motion vector detection circuit 20, the data from the small detection circuit is selected by the mode selection circuit 21, and the information is calculated by the motion compensation prediction circuit 14.
You can send it to.

In this embodiment, MPEG2 is used.
It does not mention all of the various P-picture or B-picture motion estimation schemes in the above, but it goes without saying that the motion estimation scheme does not change the essence of the present invention and does not limit the claimed scope.

According to the above embodiment, in the case of forward prediction, a motion vector detection circuit (second motion detection circuit 20 in FIG. 1) for backward prediction which has not been used conventionally is expanded. Since it is used as a motion vector detection circuit for forward prediction with respect to the search range, the accuracy of the motion vector can be increased.

Therefore, it is possible to suppress the deterioration of the decoded image quality that occurs when a special image having a fast motion is input, without significantly increasing the cost, and the image code for program material transmission that requires particularly high image quality. It is a great advantage to use it in a chemical conversion device.

(Correspondence between Invention and Embodiment) In the present embodiment, the first image memory 16 corresponds to the first image memory of the present invention, and the second image memory 17 corresponds to the second image memory of the present invention.
And the third image memory 18 corresponds to the third image memory of the present invention. Further, the first motion vector detection circuit 19 corresponds to the first motion vector detection means of the present invention, and the second motion vector detection circuit 20 corresponds to the second motion vector detection means of the present invention.

The above is the correspondence relationship between each configuration of the embodiments and each configuration of the present invention, but the present invention is not limited to the configurations of these embodiments, and the functions shown in the claims,
Needless to say, any configuration may be used as long as the functions of the embodiment can be achieved.

[0039]

As described above, according to the present invention,
In the case of bidirectional prediction, the first motion vector detection means
The motion vector from the reference image in the forward direction is detected by referring to the image data in the first image memory, and the second motion vector detecting means is used in the backward direction by referring to the image data in the second image memory. In the case of forward prediction, the first motion vector detection means operates in the same manner as in bidirectional prediction,
The second motion vector detecting means detects the motion vector from the reference image expanded in the forward direction by referring to the image data in the third image memory.

Therefore, without significantly increasing the cost,
It is possible to provide an image coding apparatus capable of improving the image quality by reducing the deterioration of the decoded image quality that occurs when a special image is input.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an image coding apparatus according to a first embodiment of the present invention.

FIG. 2 is a diagram showing a range of image data stored in each image memory of FIG.

FIG. 3 is a block diagram showing a configuration of a conventional image encoding device.

[Explanation of Codes] 2 Screen rearrangement circuit 4 DCT circuit 5 Quantization circuit 11 Inverse quantization circuit 12 Inverse DCT circuit 14 Motion compensation prediction circuit 16-18 Image memory 19, 20 Motion detection circuit 21 Mode selection circuit

Claims (5)

[Claims] 1. An image encoding apparatus for encoding a moving image using bidirectional predictive encoding, comprising: a first image memory for accumulating a forward reference image;
A first motion vector detecting means for reading out the image data of the first image memory and detecting a motion vector, a second image memory for accumulating a backward reference image, and an extended forward reference image. And a second image memory for accumulating data, and a second image memory for reading the image data of the second image memory or the third image memory and detecting a motion vector.
And a motion vector detecting means for the image coding device. 2. The second motion vector detecting means detects a motion vector by reading out image data of the second image memory in the case of bidirectional prediction, and the second motion vector detecting means in the case of forward prediction. The image coding apparatus according to claim 1, wherein the motion vector is detected by reading the image data of the image memory No. 3. 3. A selection means for selecting a motion vector detected by the first motion detection means and a motion vector detected by the second motion detection means. Image encoding device. 4. The image coding apparatus according to claim 3, further comprising a coding unit that performs motion compensation predictive coding on the image data input using the motion vector selected by the selecting unit. 5. The selecting means selects according to each prediction error signal obtained based on the motion vector detected by the first motion detecting means and the motion vector detected by the second motion detecting means. The image coding apparatus according to claim 4, wherein