JPH09261655A - Moving image coder - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain the moving image coder in which image quality of a decoded image is improved at low rate coding and detection accuracy of a motion vector is maintained even when image quality of the decoded image is deteriorated to generate a predicted image with a small prediction error. SOLUTION: The coder is provided with a frame memory 1a storing plural input images, a 1st motion vector detector 10b using the plural images stored in the frame memory 1a to detect a motion of the images, and a 2nd motion vector detector 10c correcting a motion vector outputted from the 1st motion vector detector 10b, and generates a predicted image by using the 2nd motion vector. Furthermore, a DCT transformation coefficient is used to obtain an evaluation indicating a difference between the predicted image and the input image so as to discriminate a valid block.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a moving picture coding apparatus for coding a moving picture by utilizing correlation in the time axis direction.

[0002]

2. Description of the Related Art FIG. 7 is a block diagram showing a conventional moving picture coding apparatus shown in, for example, "Latest MPEG Textbook" (ASCII Publishing Bureau, p.129-155). In FIG. 7, 1 is a frame memory for storing an input image, and the output of the frame memory 1 is given to the first input of the subtractor 2. The output of the subtractor 2 is given to the first input of the variable length encoder 5 via the DCT converter 3 and the quantizer 4. The output of the quantizer 4 is the inverse quantizer 6 and the inverse D.
It is also applied to the first input of the adder 8 via the CT converter 7. The output of the adder 8 is input to the frame memory 9.

The output of the frame memory 9 is given to the first input of the motion vector detector 10 and the first input of the predicted image forming circuit 11. The output of the frame memory 1 is given to the second input of the motion vector detector 10. The output of the motion vector detector 10 is given to the second input of the predictive image creating circuit 11 and the second input of the variable length encoder 5. The first output of the prediction image creating circuit 11 is given to the second input of the subtractor 2 and the second input of the adder 8. The second output of the prediction image creating circuit 11 is given to the third input of the variable length encoder 5. The bit stream output from the variable length encoder 5 is output via the buffer 12.

Next, the operation will be described. Motion compensation predictive coding is widely used as a coding method for reducing redundancy in the time direction of a moving picture, and the moving picture coding apparatus shown in FIG. 7 also uses the motion compensation predictive coding method. is there. FIG. 8 shows a conceptual diagram of the motion compensation predictive coding method. The input image is divided into, for example, macroblocks of 16 pixels × 16 lines, the decoded image of the image already encoded is used as a reference image, the region of the image closest to each macroblock is searched from the reference image, and the predicted image is obtained. While being created, the position of the predicted image is encoded as a motion vector. FIG. 8A shows a case of unidirectional prediction, and FIG.
(B) shows the case of bidirectional prediction.

In the case of unidirectional prediction, the reference picture is an already coded picture, which is a picture temporally earlier than the input picture.
In the case of bidirectional prediction, the reference image is an image that has already been encoded, and both an image temporally preceding and an image temporally preceding the input image are used. That is, it is necessary to change the order of the input images before encoding. In this case, the predicted image is the predicted image when the temporally previous image is the reference image, the predicted image when the temporally later image is the reference image, and the average of these predicted images. Of the three images, the one with less predictive distortion is selected.

The difference between the input image and the predicted image is coded as a prediction error using a DCT converter or the like. In addition,
An image for which unidirectional prediction is performed is called a P picture, and an image for which bidirectional prediction is performed is called a B picture. In addition, due to an initial state of encoding or a case where a transmission error occurs, encoding may be performed in a screen without performing prediction. This image is called an I picture.

The operation of the conventional moving picture coding apparatus will be described in detail below with reference to FIG. Frame memory 1
Stores the input moving image and outputs it in the encoding order. For example, when four images are input and the first image is encoded as an I picture, the second and third images are encoded as B pictures, and the fourth image is encoded as a P picture, the frame memory 1 outputs the first image first, then outputs the fourth image, and then outputs the second and third images in order.

The image output from the frame memory 1 is
The subtracter 2 obtains the difference from the predicted image output from the predicted image creation circuit 11. The prediction error output from the subtractor 2 is DCT-transformed by the DCT transformer 3 for each block of 8 pixels × 8 lines, and the transform coefficient is quantized by the quantizer 4. The quantized transform coefficient is input to the variable length encoder 5. On the other hand, the quantized transform coefficient is decoded by the inverse quantizer 6 and the inverse DCT transformer 7. The decoded prediction error is added to the predicted image output from the predicted image creation circuit 11 in the adder 8. The image output from the adder 8 is a decoded image and is input to the frame memory 9.

The predictive image is created by the motion vector detector 10 and the predictive image creating circuit 11. First, the motion vector detector 10 detects the motion of the image output from the frame memory 1 using the decoded image of the already encoded image stored in the frame memory 9 as a reference image. That is, the image output from the frame memory 1 is divided into macroblocks of 16 pixels × 16 lines to obtain the motion vector of each macroblock. At this time, when the input image is a frame composed of two fields of interlace, a motion vector for 16 pixels × 16 lines of the frame is obtained, and this macroblock is divided into 16 pixels × 8 lines of two fields. The motion vectors of the two fields are also obtained.

The predicted image creating circuit 11 creates a predicted image from the decoded image stored in the frame memory 9 based on the motion vector output from the motion vector detector 10. For example, when the motion vector output from the motion vector detector 10 indicates the position of a half pixel,
The pixel value at the position of a half pixel is calculated from the surrounding four pixels. When the input image is a frame composed of two interlaced fields, a prediction image is created for each of the frame motion vector and the field motion vector, and the one with the smaller prediction distortion is selected and output. To do.

When the image output from the frame memory 1 is a B-picture, two predicted images are created from two reference images, and an average image of these two predicted images is created. Of the two prediction images, the one with less prediction distortion is selected. Further, when performing intra-image coding, all pixel values of the predicted image are set to 0 and output. The prediction image creation circuit 11 creates and outputs the prediction image as described above, and also outputs the prediction mode indicating which prediction image is selected.

The variable-length encoder 5 includes a quantized transform coefficient output from the quantizer 4, a prediction mode output from the prediction image creating circuit 11, and a motion vector output from the motion vector detector 10. Variable-length encoding
It is superimposed and output as a bit stream via the buffer 12.

[0013]

In the conventional moving picture coding apparatus, in order to perform coding at a low rate of, for example, about 2 Mbit / s, it is necessary to increase the quantization step width of the quantizer. There has been a problem that the quantization error becomes large and the quality of the decoded image is greatly deteriorated as a whole.

Further, in the conventional moving picture coding apparatus, since the decoded picture of the already coded picture is used for the detection of the motion vector, the movement of the motion picture is deteriorated due to the deterioration of the picture quality of the decoded picture, especially at the low rate coding. There is a problem that the detection accuracy of the vector is reduced.

The present invention has been made for the purpose of solving the above problems, and it is possible to improve the image quality of a decoded image in low-rate encoding, and even if the image quality of the decoded image deteriorates, the motion vector An object of the present invention is to obtain a moving picture coding device which can maintain a detection accuracy and can obtain a predicted image with a small prediction error.

[0016]

According to a first aspect of the present invention, a moving picture coding apparatus divides an input image into areas having a plurality of pixels, and detects a motion vector indicating the movement of the moving picture for each area. A vector detection unit, a prediction image generation unit that generates a motion-compensated prediction image from a decoded image of an image that has already been encoded based on the detected motion vector, and a difference signal between the input image and the prediction image. A subtractor,
This difference signal is divided into blocks for each of a plurality of pixels to obtain an evaluation value indicating the magnitude of the low-frequency component of each block, and if this evaluation value is within the threshold range, it is determined that the block is an invalid block. Further, it is provided with a valid block determination unit which determines that the block is a valid block when the block is outside the threshold range, and an encoder which encodes the block determined to be the valid block.

A moving picture coding apparatus according to a second invention is the moving picture coding apparatus according to the first invention, further comprising means for controlling a threshold value based on the code amount output from the encoder. .

A moving picture coding apparatus according to a third aspect of the present invention includes a motion vector detecting section that divides an input image into areas consisting of a plurality of pixels and detects a motion vector indicating the motion of the moving picture for each area. A prediction image generation unit that generates a motion-compensated prediction image from a decoded image of an image that has already been encoded based on the detected motion vector, a subtractor that generates a difference signal between the input image and the prediction image, and this difference An orthogonal transform circuit for performing a orthogonal transform for each block by grouping signals into a plurality of pixels and determining the magnitude of the low-frequency component of the difference signal from the low-frequency coefficient among the transform coefficients output from this orthogonal transform circuit. The evaluation value shown is determined, and when this evaluation value is within the threshold range, it is determined that the block is an invalid block, and when it is outside the threshold range, the block is a valid block. An effective block determination unit that determines that,
An encoder that encodes the transform coefficient output from the orthogonal transform circuit for the block determined to be the effective block.

A moving picture coding apparatus according to a fourth aspect of the present invention includes a motion vector detection unit that divides an input image into areas having a plurality of pixels and detects a plurality of motion vectors indicating the movement of the moving picture for each area. , Prediction that outputs a prediction image that gives the minimum prediction distortion by creating a plurality of motion-compensated prediction images from the decoded image of the image already encoded based on the detected motion vector An image creating unit, a subtracter that generates a difference signal between the input image and the predicted image, and if the minimum prediction distortion is within a threshold range, it is determined that the area is an invalid block, and An effective block determination unit that determines that the area is an effective block when the area is out of the range, and an encoder that encodes the difference signal for the area determined to be the effective block A.

A moving picture coding apparatus according to a fifth invention is the moving picture coding apparatus according to the fourth invention, further comprising means for controlling a threshold value based on the code amount output from the encoder. is there.

According to a sixth aspect of the present invention, there is provided a moving picture coding apparatus which includes a frame memory for storing a plurality of input images and a plurality of motion vectors indicating the motion of the moving picture from the plurality of images stored in the frame memory. A motion vector detection unit for detecting and a plurality of motion-compensated prediction images are created from decoded images of images already coded based on the detected motion vectors, the prediction distortion of each prediction image is obtained, and the minimum prediction distortion is calculated. A prediction image creation unit that outputs a given prediction image and a subtractor that generates a difference signal between the input image and the prediction image are provided.

A moving picture coding apparatus according to a seventh aspect of the present invention detects a frame memory for storing a plurality of input images and a motion vector indicating the motion of the moving picture from the plurality of images stored in the frame memory. A first motion vector detection unit, and a second motion vector detection that performs a motion vector search using a decoded image of an already encoded image with the motion vector output from the first motion vector detection unit as the center. Unit, a prediction image creating unit that creates a motion-compensated prediction image from a decoded image of an image already encoded based on the motion vector output from the second motion vector detection unit, the prediction image and the input image And a subtractor for generating the differential signal of.

A moving picture coding apparatus according to an eighth aspect of the present invention includes a frame memory for storing a plurality of input images and a plurality of motion vectors indicating a motion of the moving picture from the plurality of images stored in the frame memory. A motion vector detection unit for detecting and a plurality of motion-compensated prediction images are created from decoded images of images already coded based on the detected motion vectors to obtain prediction distortion of each prediction image, and the minimum prediction distortion is calculated. A prediction image generation unit that outputs a given prediction image, a subtracter that generates a difference signal between the input image and the prediction image, and a block of the difference signal for each plurality of pixels to determine the magnitude of the difference signal of each block. If the evaluation value is within the threshold range, it is determined that the block is an invalid block.If the evaluation value is outside the threshold range, the block is valid. An effective block determination unit to determine that Tsu is click, the valid block determined to be the block is obtained by a coder for encoding.

A moving picture coding apparatus according to a ninth aspect of the invention detects a frame memory for storing a plurality of input images and a motion vector indicating the motion of the moving picture from the plurality of images stored in the frame memory. A first motion vector detection unit, and a second motion vector detection that performs a motion vector search using a decoded image of an already encoded image with the motion vector output from the first motion vector detection unit as the center. Unit, a prediction image creation unit that creates a motion-compensated prediction image from a decoded image of an image already encoded based on the motion vector output from the second motion vector detection unit, the input image and the prediction image And a subtractor for generating the difference signal of each block, and the difference signal is divided into blocks for each of a plurality of pixels to obtain an evaluation value indicating the magnitude of the difference signal of each block. If it is within the range, it is determined that the block is an invalid block, and if it is outside the threshold range, a valid block determination unit that determines that the block is a valid block, and the valid block is determined. And an encoder for encoding the blocks.

[0025]

BEST MODE FOR CARRYING OUT THE INVENTION A moving picture coding apparatus according to a first invention divides an input image into areas each having a plurality of pixels, detects a motion vector indicating a motion of the moving picture for each area, and detects the motion vector. A motion-compensated predicted image is created from a decoded image of an image already encoded based on a vector, a difference signal between the input image and the predicted image is generated, and the difference signal is divided into blocks for each of a plurality of pixels to obtain a block of each block. Obtaining an evaluation value indicating the magnitude of the low-frequency component, if this evaluation value is within the range of the threshold value, it is determined that the block is an invalid block, and only the information indicating the invalid block is encoded, and If it is outside the threshold range, it is determined that the block is a valid block and the block is encoded.

In the moving picture coding apparatus according to the second invention, in the moving picture coding apparatus according to the first invention, the means for controlling the threshold value controls the threshold value on the basis of the generated code amount, and the effective block Controls the number of blocks determined.

The moving picture coding apparatus according to the third invention divides the input picture into areas consisting of a plurality of pixels, detects a motion vector indicating the motion of the moving picture for each area, and based on this motion vector, A motion-compensated predicted image is created from a decoded image of the coded image, a differential signal between the input image and the predicted image is generated, and the differential signal is grouped into a plurality of pixels into blocks and orthogonally transformed for each block. To obtain the conversion coefficient, obtain an evaluation value indicating the magnitude of the low frequency component of the difference signal from the low frequency coefficient among these conversion coefficients, and if this evaluation value is within the threshold range, the block is Only the information indicating that the block is an invalid block is coded, and if the evaluation value is outside the range of the threshold, the block is determined to be a valid block and the block is blocked. Encoding the transform coefficients.

A moving picture coding apparatus according to a fourth aspect of the invention divides an input picture into areas consisting of a plurality of pixels, detects a plurality of motion vectors indicating the movement of the moving picture for each area, and uses this motion vector as the motion vector. Based on the decoded image of the image already encoded based on a plurality of motion-compensated predicted image to create a prediction distortion of each predicted image, to generate a difference signal between the predicted image and the input image to give the minimum prediction distortion, If the minimum prediction distortion is within the range of the threshold value, it is determined that the area is an invalid block and only the information indicating that the area is an invalid block is coded. It is determined that the block is a valid block, and the differential signal of the area is encoded.

In the moving picture coding apparatus according to the fifth invention, in the moving picture coding apparatus according to the fourth invention, the means for controlling the threshold value controls the threshold value based on the generated code amount, and the effective block Controls the number of blocks determined.

A moving picture coding apparatus according to a sixth invention comprises a frame memory for storing a plurality of input images, and a plurality of motion vectors indicating the motion of the moving picture from the plurality of images stored in the frame memory. Is detected, a plurality of motion-compensated prediction images are created from the decoded image of the image already encoded based on this motion vector, the prediction distortion of each prediction image is obtained, and a prediction image that gives the minimum prediction distortion is obtained, A difference signal from the input image is generated.

The moving picture coding apparatus according to the seventh invention comprises a frame memory for storing a plurality of input images, and detects a motion vector indicating the motion of the moving picture from the plurality of images stored in the frame memory. Then, the motion vector is searched by using the decoded image of the image already encoded around this motion vector to correct the motion vector, and based on this corrected motion vector, from the decoded image of the image already encoded. A motion-compensated predicted image is created, and a difference signal between this predicted image and the input image is generated.

The moving picture coding apparatus according to the eighth invention comprises a frame memory for storing a plurality of input images, and a plurality of motion vectors indicating the motion of the moving picture from the plurality of images stored in the frame memory. Is detected, a plurality of motion-compensated prediction images are created from decoded images of images already encoded based on this motion vector, the prediction distortion of each prediction image is obtained, and a prediction image that gives the minimum prediction distortion is obtained. When a difference signal between the predicted image and the input image is generated, the difference signal is divided into blocks for each of a plurality of pixels to obtain an evaluation value indicating the magnitude of the difference signal of each block, and when the evaluation value is within a threshold range Determines that the block is an invalid block and encodes only the information indicating that the block is an invalid block. If the block is outside the threshold range, the block is a valid block. Determined for coding of the blocks.

The moving picture coding apparatus according to the ninth invention comprises a frame memory for storing a plurality of input images, and detects a motion vector indicating the motion of the moving picture from the plurality of images stored in the frame memory. Then, the motion vector is searched by using the decoded image of the image already encoded around this motion vector to correct the motion vector, and the motion is performed from the decoded image of the image already encoded based on this corrected motion vector. A compensated predicted image is created, a difference signal between the predicted image and the input image is generated, and the difference signal is divided into blocks for each of a plurality of pixels to obtain an evaluation value indicating the magnitude of the difference signal of each block. If the evaluation value is within the threshold range, it is determined that the block is an invalid block and only the information indicating the invalid block is coded. Performing encoding of the block is determined as the block is valid block if it is.

Hereinafter, the present invention will be specifically described with reference to the drawings showing the embodiments. Embodiment 1. 1 is a block diagram showing a moving picture coding apparatus according to a first embodiment of the present invention. The same reference numerals as those in FIG. 7 denote the same or corresponding portions. In the figure, 13a is a valid block determiner, and the output of the DCT converter 3 is given to the first input of the valid block determiner 13a and the first input of the quantizer 4a. Quantizer 4a
Output of the variable length encoder 5 and the inverse quantizer 6
, And the output of the variable-length encoder 5 is supplied to the input of the buffer 12a. The first output of the buffer 12a is transmitted or stored to the outside, and the second output of the buffer 12a is given to the second input of the valid block determiner 13a. The output of the valid block determiner 13a is given to the second input of the quantizer 4a. The quantizer 4a and the variable-length encoder 5 constitute an encoder 14 that encodes the output of the DCT converter 3.

Next, the operation will be described. The moving picture coding apparatus according to the first embodiment is also a coding apparatus using the motion compensation predictive coding method. First, the frame memory 1 stores the input moving image and outputs it in the encoding order. That is, as described in the conventional example, when using bidirectional prediction, it is necessary to change the order of input images before encoding. Therefore, for example, when four images are input, the first image is encoded as an I picture, the second and third images are encoded as B pictures, and the fourth image is encoded as a P picture, The frame memory 1 outputs the first image first, then the fourth image, and then the second and third images in order.

The image output from the frame memory 1 is
The subtracter 2 obtains the difference from the predicted image output from the predicted image creation circuit 11. The prediction error output from the subtractor 2 is divided into blocks of 8 pixels × 8 lines in the DCT converter 3, and DCT conversion is performed for each block. The effective block determiner 13a uses this transform coefficient to determine whether or not to encode each block.

That is, in low-rate coding,
Since exact restoration of high frequency components is not required, the subtractor 2
If the prediction error output from the block includes only high-frequency components (that is, does not include low-frequency components), even if the pixel value in the block is replaced with 0 as an invalid block, the image quality deterioration in the decoded image Stays within the acceptable range. Further, when the block is an invalid block, the transform coefficient of the block is not encoded and only the information that the block is an invalid block is encoded. Therefore, the amount of code can be reduced. The image quality of the effective block can be improved.

The effective block determiner 13a uses the transform coefficient to determine whether each block contains a low frequency component. That is, when DCT transform is applied to a block of 8 pixels × 8 lines, 8 × 8 transform coefficients t (i, j) (0 ≦ i, j <8) as shown in FIG. 2 are output, and each transform coefficient is output. Corresponds to the frequency component. t (0,0) represents a direct current component (DC), t (7,0) represents a high frequency component in the horizontal direction, t (0,7) represents a high frequency component in the vertical direction, and t (7,7). ) Represents a high frequency component both horizontally and vertically. The other transform coefficients respectively correspond to the intermediate frequency components. Therefore, the conversion coefficient t (0,0) representing the DC component is particularly called a DC coefficient.

In general, the transform coefficient can be divided into a low frequency coefficient and a high frequency coefficient according to the corresponding frequency. Therefore, the magnitude of the low frequency component of each block can be easily evaluated by using the low frequency coefficient representing the low frequency component. For example, if the absolute value of the DC coefficient t (0,0) is greater than or equal to the threshold value, it is determined that the low frequency component is included and the block is determined to be a valid block. If the absolute value of the DC coefficient is less than the threshold value, the invalid block is determined. To determine.

The quantizer 4a quantizes the transform coefficient output from the DCT transformer 3. At this time, for the blocks determined to be invalid blocks by the valid block determiner 13a, all the transform coefficient values are replaced with 0. The quantized transform coefficient is input to the variable length encoder 5. The variable-length encoder 5 encodes the information that the block is an invalid block when the values of the quantized transform coefficients are all 0, and otherwise the block is an effective block. The information and each transform coefficient are encoded.

On the other hand, the quantized transform coefficient is decoded by the inverse quantizer 6 and the inverse DCT transformer 7. This decoded prediction error is added to the predicted image creation circuit 1 in the adder 8.
The predicted image output from 1 is added. The image output from the adder 8 is a decoded image and is input to the frame memory 9. The operations of the motion vector detector 10 and the prediction image creation circuit 11 are exactly the same as those of the conventional example, and thus the description thereof will be omitted.

The prediction mode output from the predictive image generation circuit 11 and the motion vector output from the motion vector detector 10 are input to the variable length encoder 5 and encoded.
The variable-length encoder 5 superimposes these coded data and the coded data of the above-mentioned transform coefficient, and outputs them as a bit stream to the buffer 12a. The buffer 12a outputs this bit stream to the outside and also outputs the code amount in the buffer to the valid block determiner 13a. The valid block determiner 13a changes the threshold value based on the code amount in the buffer to control the number of valid blocks to be coded.

In the first embodiment, the DCT transform is used for coding the prediction error, but other orthogonal transforms such as KL transform and Hadamard transform, subband coding, wavelet coding, etc. are used. Also, similarly, the effective block can be determined using the low-frequency coefficient. Further, as the low-frequency coefficient, not only the DC coefficient t (0,0) but also a plurality of low-frequency coefficients are used to obtain an evaluation value indicating the magnitude of the low-frequency component, and the effective value is obtained based on this evaluation value. It may be configured to determine the block.

For example, as the low frequency coefficient, t (n, m)
Select (0 ≦ n, m <4),

[0045]

[Equation 1]

The evaluation value d1 is obtained by the following. If d1 is within the threshold range, the block is judged as an invalid block,
If it is outside the range of the threshold value, the block may be determined to be an effective block.

As the low-frequency coefficient, t (n, m) (0
≦ n, m <8),

[0048]

[Equation 2]

The evaluation value d2 is obtained by the following. If d2 is within the range of the threshold value, the block is judged as an invalid block,
If it is outside the range of the threshold value, the block may be determined to be an effective block.

Alternatively, the evaluation value d3 may be obtained as d3 = max {| t (n, m) |; 0 ≦ n, m <8}, and the determination may be made based on this evaluation value d3. Good.

Further, even when a prediction error is coded by a coding method that does not perform division in the frequency direction, such as DPCM or vector quantization, the prediction error is reduced by passing it through a low-pass filter. It is possible to obtain a band frequency component, obtain an evaluation value indicating the magnitude of the low band frequency component for each block, and determine an effective block based on this evaluation value.

Further, in the first embodiment, the motion vector detector 10 and the predicted image creating circuit 11 operate in the same manner as in the conventional example, in which the motion vector is obtained in macroblock units of 16 pixels × 16 lines. The compensation method is not limited to this, and for example, a motion vector may be obtained for each area of an arbitrary shape. In particular, the macroblock may be divided according to the prediction error, and the motion vector may be obtained for each variable-sized block. Further, it may be configured such that a region is divided for each moving object in the image and a motion vector is obtained for each region.

Embodiment 2 3 is a block diagram showing a moving picture coding apparatus according to a second embodiment of the present invention. The same reference numerals as those in FIG. 1 denote the same or corresponding portions. In the figure, 1 is a frame memory for storing an input image, and the output of the frame memory 1 is given to a first input of a subtractor 2. The output of the subtractor 2 is given to the first input of the quantizer 4b via the DCT converter 3. The output of the effective block determiner 13b is given to the second input of the quantizer 4b.

The output of the quantizer 4b is given to the first input of the variable length encoder 5 and the input of the inverse quantizer 6. The output of the inverse quantizer 6 is given to the first input of the adder 8 via the inverse DCT converter 7. The output of the adder 8 is input to the frame memory 9. The output of the frame memory 9 is the first input of the motion vector detector 10 and the predicted image generation circuit 1
Given to the first input of 1a.

The output of the frame memory 1 is applied to the second input of the motion vector detector 10. The output of the motion vector detector 10 is given to the second input of the prediction image creating circuit 11a and the second input of the variable length encoder 5. The first output of the prediction image creation circuit 11a is given to the second input of the subtractor 2 and the second input of the adder 8. The second output of the predictive image creating circuit 11a is given to the third input of the variable length encoder 5, and the third output of the predictive image creating circuit 11a is given to the first input of the effective block determiner 13b.

The bit stream output from the variable length encoder 5 is input to the buffer 12a. Buffer 12
The first output of a is transmitted or stored externally, and the second output of the buffer 12a is given to the second input of the valid block determiner 13b. The DCT converter 3, the quantizer 4b, and the variable-length encoder 5 form an encoder 15 that encodes the difference signal output from the subtractor 2.

Next, the operation will be described. The second embodiment differs from the first embodiment in that the evaluation value used for the determination by the valid block determiner 13b is the prediction distortion output from the prediction image creation circuit 11a. First, the frame memory 1 stores the input moving image and outputs it in the encoding order. The image output from the frame memory 1 is the subtractor 2
At, the difference from the predicted image output from the predicted image creation circuit 11a is obtained. The prediction error output from the subtractor 2 is divided into blocks in the DCT converter 3, and DCT-transformed for each block.

The DCT-transformed transform coefficient is the quantizer 4b.
Is quantized by. At this time, the effective block determiner 13b
For the blocks in the area determined to be invalid blocks by, all the transform coefficient values are replaced with 0. The quantized transform coefficient is input to the variable length encoder 5. The variable-length encoder 5 encodes the information that the block is an invalid block when all the quantized transform coefficient values are 0, and the information that the block is an effective block otherwise. And each transform coefficient is encoded.

On the other hand, the quantized transform coefficient is decoded by the inverse quantizer 6 and the inverse DCT transformer 7. The decoded prediction error is added to the predicted image output from the predicted image creation circuit 11a in the adder 8. Adder 8
The image output from is a decoded image and is input to the frame memory 9.

The predicted image is created by the motion vector detector 10 and the predicted image creating circuit 11a. First, the motion vector detector 10 detects the motion of the image output from the frame memory 1 using the decoded image of the already encoded image stored in the frame memory 9 as a reference image.
For example, an image output from the frame memory 1
It is divided into macroblocks of pixels × 16 lines, and the motion vector of each macroblock is obtained.

At this time, the motion vector for 16 pixels × 16 lines of the frame is obtained, and this macroblock is divided into 16 pixels × 8 lines of two fields to obtain the motion vectors of the two fields. Further, when the image output from the frame memory 1 is a B picture, the two images stored in the frame memory 9 are used as reference images, and the motion vectors of the frame and the field are obtained.

The predictive image creating circuit 11a creates a plurality of predictive images from the decoded images stored in the frame memory 9 based on the plurality of motion vectors output from the motion vector detector 10 and calculates each predictive image. Find the prediction distortion of
A predicted image that gives the minimum predicted distortion is obtained. For example,
In the case of P picture, a prediction image is created for each of the frame motion vector and the field motion vector,
The one with the smaller predicted distortion is selected and output.

In the case of a B picture, two prediction images are created from two reference images for the frame motion vector and the field motion vector, respectively, and an average image of these two prediction images is created. , Total 3
One of the predicted images having a small prediction distortion is selected, and further, of the predicted images selected in each of the frame and the field, one having a small prediction distortion is selected. As the prediction distortion, sum of absolute values of prediction errors, sum of squares, or
A value obtained by adding an offset value depending on the magnitude of the motion vector to these is used.

When the motion vector output from the motion vector detector 10 indicates the position of a half pixel,
The pixel value at the position of a half pixel is calculated from the surrounding four pixels. Further, when performing intra-image coding, all pixel values of the predicted image are set to 0 and output. Prediction image creation circuit 11a
Generates and outputs the prediction image as described above, and outputs the prediction mode indicating which prediction image is selected and the prediction distortion when the prediction image is selected.

The prediction mode output from the predictive image generation circuit 11a and the motion vector output from the motion vector detector 10 are input to the variable length encoder 5 and encoded. The variable length encoder 5 superimposes these coded data and the coded data of the above-mentioned transform coefficient, and outputs it as a bit stream to the buffer 12a. The buffer 12a is
This bit stream is output to the outside and the code amount in the buffer is output to the valid block determiner 13b.

On the other hand, the predictive distortion output from the predictive image creating circuit 11a is input to the effective block determiner 13b. When the value of the prediction distortion is equal to or less than the threshold, the valid block determiner 13b determines that the macroblock is the same as the predicted image, determines the macroblock as an invalid block, and the value of the prediction distortion is greater than the threshold. When it is larger, the macro block is determined to be a valid block. further,
The valid block determiner 13b changes the threshold value based on the code amount in the buffer 12a to control the number of valid blocks to be coded.

In the second embodiment, the motion vector detector 10 and the prediction image generation circuit 11a calculate a motion vector in units of macroblocks of 16 pixels × 16 lines to generate a prediction image. The method is not limited to this, and for example, a motion vector may be obtained for each area of any shape. In particular, the macroblock may be divided according to the prediction error, and the motion vector may be obtained for each variable-sized block. Further, it may be configured such that a region is divided for each moving object in the image and a motion vector is obtained for each region. The coding of the prediction error is not limited to DCT, and any coding method can be used.

Embodiment 3 4 is a block diagram showing a moving picture coding apparatus according to a third embodiment of the present invention. The same reference numerals as those in FIG. 7 denote the same or corresponding portions. In the figure, 1a is a frame memory for storing an input image, and the first output of the frame memory 1a is given to the first input of the subtractor 2 and the first input of the motion vector detector 10a, The second output of 1a is provided to the second input of the motion vector detector 10a. The output of the subtractor 2 is given to the first input of the variable length encoder 5 via the DCT converter 3 and the quantizer 4.

The output of the quantizer 4 is also given to the first input of the adder 8 via the inverse quantizer 6 and the inverse DCT converter 7. The output of the adder 8 is input to the frame memory 9. The output of the frame memory 9 is given to the first input of the prediction image creating circuit 11. Prediction image creation circuit 1
The output of the motion vector detector 10a is given to the second input of 1. The output of the motion vector detector 10a is also given to the second input of the variable length encoder 5. The first output of the prediction image creating circuit 11 is given to the second input of the subtractor 2 and the second input of the adder 8. Prediction image creation circuit 1
The second output of 1 is provided to the third input of the variable length encoder 5. The bit stream output from the variable length encoder 5 is output via the buffer 12.

Next, the operation will be described. Embodiment 3
Is different from the conventional example in the method of detecting the motion vector. Since the motion vector detection of the conventional example uses the decoded image of the already encoded image, there is a problem that the motion vector detection accuracy is deteriorated due to the deterioration of the image quality of the decoded image. On the other hand, if a motion vector is detected in an input image with no image quality deterioration, the motion can be grasped faithfully, but the decoded image is used to create the predicted image. The optimum motion vector is not always selected in the sense of “reducing
Therefore, in the third embodiment, a plurality of motion vectors are detected using the input image, a plurality of prediction images are created using decoded images based on these motion vectors, and the prediction image with the smallest prediction distortion is generated. Select.

A detailed description will be given below with reference to FIG. First, the frame memory 1a stores an input moving image,
The data is rearranged in the encoding order and is output from the first output to the subtractor 2. The subtractor 2 obtains the difference between the image output from the frame memory 1a and the predicted image output from the predicted image creation circuit 11. The prediction error output from the subtractor 2 is DCT-transformed by the DCT transformer 3 for each block of 8 pixels × 8 lines, and the transform coefficient thereof is quantized by the quantizer 4. The quantized transform coefficient is input to the variable length encoder 5. On the other hand, the quantized transform coefficient is decoded by the inverse quantizer 6 and the inverse DCT transformer 7. The decoded prediction error is added to the predicted image output from the predicted image generation circuit 11 in the adder 8. The image output from the adder 8 is a decoded image and is input to the frame memory 9.

The predicted image is created by the motion vector detector 10a and the predicted image creating circuit 11. First, the motion vector detector 10a detects the motion of the image output from the first output of the frame memory 1a, using the input image stored in the frame memory 1a as a reference image. For example, the image output from the first output of the frame memory 1a is divided into macroblocks of 16 pixels × 16 lines to obtain the motion vector of each macroblock. At this time, the motion vector for 16 pixels × 16 lines of the frame is obtained, and this macroblock is divided into 16 pixels × 8 lines of two fields to obtain the motion vectors of the two fields. Also, the frame memory 1
When the image output from the first output of a is a B picture, the two images stored in the frame memory 1a are used as reference images, and the motion vectors of the frame and the field are obtained.

The predicted image creating circuit 11 creates a plurality of predicted images from the decoded images stored in the frame memory 9 based on the plurality of motion vectors output from the motion vector detector 10a, and creates each predicted image. Find the prediction distortion of
A predicted image that gives the minimum predicted distortion is obtained. For example,
In the case of P picture, a prediction image is created for each of the frame motion vector and the field motion vector,
The one with the smaller predicted distortion is selected and output. Further, in the case of a B picture, two prediction images are created from two reference images for each of the frame motion vector and the field motion vector, and an average image of the two prediction images is created. One of the predicted images having a small prediction distortion is selected, and further, of the predicted images selected in each of the frame and the field, one having a small prediction distortion is selected.

When the motion vector output from the motion vector detector 10a indicates the position of a half pixel, the pixel value at the position of the half pixel is calculated from the surrounding four pixels. Further, when performing intra-image coding, all pixel values of the predicted image are set to 0 and output. The prediction image creation circuit 11 creates and outputs the prediction image as described above, and also outputs the prediction mode indicating which prediction image is selected.

The variable length encoder 5 includes the quantized transform coefficient output from the quantizer 4, the prediction mode output from the predictive image creating circuit 11, and the motion vector detector 10a.
Of the motion vectors output from the above are respectively variable length coded and superimposed, and output via the buffer 12 as a bit stream.

As described above, according to the third embodiment, the predictive image with the smallest predictive distortion is selected from the plurality of motion vectors output from the motion vector detector 10a. Can be improved. In particular, in a B picture, there is a prediction mode called an average image of two predicted images created from two reference images. In this prediction mode, deterioration of image quality in a decoded image can be reduced by a filter called an averaging process. In order to evaluate this reduction effect, it is very important to compare prediction distortions using decoded images.
In fact, if the selection of the prediction mode in the B picture is decided by the input image without image quality deterioration, the average image is conspicuously deteriorated in sharpness, so that it is not often selected.

On the other hand, when the determination is made using the decoded image as in the third embodiment, the above-described filter effect is exhibited, the average image is selected most, and the image quality of the predicted image is considerably improved. Further, when actually compared with the conventional moving picture coding apparatus shown in FIG. 7 at 2 Mbit / s, the S / N of the decoded picture is improved by about 0.6 dB.

In the third embodiment, the plurality of motion vectors detected by the motion vector detector 10a are
In the case of the P picture, the motion vector of the frame and the field was used, and in the case of the B picture, the motion vector of the frame and the field for each of the two reference images, but the plurality of motion vectors detected by the motion vector detector 10a are It is not limited to this. For example, 16 pixels x 16
For each of the macroblocks of the line, by providing a motion vector that gives the minimum prediction distortion and a motion vector that gives the second and third smallest prediction distortions by block matching, a greater effect is exhibited. be able to.

Fourth Embodiment FIG. 5 is a block diagram showing a moving picture coding apparatus according to the fourth embodiment of the present invention. The same reference numerals as those in FIG. 4 indicate the same or corresponding portions. In the figure, 1a is a frame memory for storing an input image, and a first output of the frame memory 1a is a first input of a subtractor 2 and a first input of a first motion vector detector 10b and a second input. To the first input of the motion vector detector 10c, and the second output of the frame memory 1a to the second input of the first motion vector detector 10b.
The output of the subtractor 2 is given to the first input of the variable length encoder 5 via the DCT converter 3 and the quantizer 4. The output of the quantizer 4 is the inverse quantizer 6 and the inverse DCT converter 7
Is also provided to the first input of the adder 8.

The output of the adder 8 is input to the frame memory 9. The output of the frame memory 9 is the second input of the second motion vector detector 10c and the prediction image creation circuit 11
Is given to the first input. The first motion vector detector 1 is connected to the third input of the second motion vector detector 10c.
The output of 0b is given. Second motion vector detector 1
The output of 0c is the second input of the prediction image creating circuit 11,
It is provided to the second input of the variable length encoder 5. The first output of the prediction image creating circuit 11 is given to the second input of the subtractor 2 and the second input of the adder 8. Prediction image creation circuit 1
The second output of 1 is provided to the third input of the variable length encoder 5. The bit stream output from the variable length encoder 5 is output via the buffer 12.

Next, the operation will be described. Embodiment 4
Is different from the conventional example in the method of detecting the motion vector. In the fourth embodiment, the first motion vector is detected using the input image, the second motion vector is searched using the decoded image centering on the first motion vector, and the prediction image generation circuit is used. The motion vector given to 11 is corrected.

A detailed description will be given below with reference to FIG. First, the frame memory 1a stores an input moving image,
The data is rearranged in the encoding order and is output from the first output to the subtractor 2. The subtractor 2 obtains the difference between the image output from the frame memory 1a and the predicted image output from the predicted image creation circuit 11. The prediction error output from the subtractor 2 is DCT-transformed by the DCT transformer 3 for each block of 8 pixels × 8 lines, and the transform coefficient is quantized by the quantizer 4. The quantized transform coefficient is input to the variable length encoder 5. On the other hand, the quantized transform coefficient is decoded by the inverse quantizer 6 and the inverse DCT transformer 7. This decoded prediction error is added to the predicted image creation circuit 1 in the adder 8.
The predicted image output from 1 is added. The image output from the adder 8 is a decoded image and is input to the frame memory 9.

The predicted image is the first motion vector detector 1
0b, the second motion vector detector 10c, and the predicted image generation circuit 11. First, the first motion vector detector 10b detects the motion of the image output from the first output of the frame memory 1a, using the input image stored in the frame memory 1a as a reference image.

For example, the image output from the first output of the frame memory 1a is divided into macroblocks of 16 pixels × 16 lines, and the motion vector of each macroblock is obtained with pixel accuracy. At this time, when the input image is a frame composed of two interlaced fields,
A motion vector for 16 pixels × 16 lines of the frame is obtained, and this macroblock is divided into 16 pixels × 8 lines of two fields to obtain the motion vectors of the two fields. When the image output from the first output of the frame memory 1a is a B picture, the motion vector is calculated for each of the two images stored in the frame memory 1a as reference images.

The second motion vector detector 10c searches for the motion vector with the motion vector output from the first motion vector detector 10b as the center and the decoded image stored in the frame memory 9 as the reference image. I do. For example, centering on the motion vector detected by the first motion vector detector 10b with pixel accuracy, a motion vector is searched from a total of nine motion vectors including this motion vector and the motion vectors at the surrounding half pixel positions.

That is, the prediction distortion is obtained for each of the nine motion vectors, and the vector with the smallest prediction distortion is output as the second motion vector. The prediction distortion at this time is obtained using the output of the frame memory 1a and the decoded image stored in the frame memory 9. When the first motion vector detector 10b outputs a plurality of motion vectors, each motion vector is centered and a search is performed to correct the motion vector.

The predicted image creating circuit 11 creates a predicted image from the decoded image stored in the frame memory 9 based on the second motion vector output from the second motion vector detector 10c. For example, when the second motion vector output from the second motion vector detector 10c indicates the position of a half pixel, the pixel value at the position of the half pixel is calculated from the surrounding four pixels. When the input image is a frame composed of two interlaced fields, a prediction image is created for each of the frame motion vector and the field motion vector, and the one with the smaller prediction distortion is selected and output. .

When the image output from the first output of the frame memory 1a is a B picture, two predicted images are created from two reference images, and the average image of these two predicted images is calculated. Created and selecting one with less prediction distortion out of a total of three predicted images. Further, when performing intra-image coding, all pixel values of the predicted image are set to 0 and output. The prediction image creation circuit 11 creates and outputs the prediction image as described above, and also outputs the prediction mode indicating which prediction image is selected.

The variable length encoder 5 outputs the quantized transform coefficient output from the quantizer 4, the prediction mode output from the prediction image creating circuit 11, and the second motion vector detector 10c. The respective second motion vectors are subjected to variable length coding, superimposed, and output as a bit stream via the buffer 12.

As described above, according to the fourth embodiment, the search for the second motion vector using the decoded image is centered on the first motion vector obtained using the input image with no image quality deterioration. Since this is performed, it is possible to select a motion vector that is close to the actual motion, and also select a motion vector that also considers the image quality deterioration of the decoded image. Actually, the S / N ratio of the decoded image was improved by about 0.5 dB when compared with the conventional moving image coding apparatus shown in FIG. 7 at 2 Mbit / s.

In the fourth embodiment, the second motion vector detector 10c uses the first motion vector detected by the first motion vector detector 10b as the center and sets the surrounding half-pixel precision vector. Although the second motion vector is searched for among the nine included vectors, the search range of the second motion vector detector 10c is not limited to this, and a wide range may be searched. In particular, in low rate coding in which the image quality deterioration of the decoded image locally increases, the search range of the second motion vector detector 10c is widened and a part of the range already searched by the first motion vector detector 10b is expanded. May be included.

In the fourth embodiment, the first motion vector detector 10b calculates the motion vector with pixel accuracy, but the accuracy of the motion vector in the first motion vector detector 10b is not limited to this. For example, the search may be performed with a precision of 2 pixels.

Further, in the third and fourth embodiments, the prediction distortion is obtained by using the sum of absolute values of prediction errors or the sum of squares, or a value obtained by adding an offset value depending on the magnitude of the motion vector to them. Good.

Embodiment 5 Since the image quality of the predicted image is improved by using the above-described Embodiment 3 or Embodiment 4,
By combining these embodiments with the above-described first or second embodiment, it is possible to obtain a moving picture coding apparatus more suitable for low rate coding. For example, FIG. 6 is a block diagram showing a moving picture coding apparatus in which the first embodiment and the fourth embodiment are combined. In the figure, the same reference numerals as those in FIGS. 1 and 5 indicate the same or corresponding portions.

In FIG. 6, reference numeral 1a is a frame memory for storing an input image, and the first output of the frame memory 1a is the first input of the subtracter 2 and the first output of the first motion vector detector 10b. Input and second motion vector detector 10
The second output of the frame memory 1a is supplied to the second input of the first motion vector detector 10b. The output of the subtractor 2 is input to the DCT converter 3. The output of the DCT converter 3 is given to the first input of the effective block determiner 13a and the first input of the quantizer 4a. The output of the effective block determiner 13a is given to the second input of the quantizer 4a. The output of the quantizer 4a is given to the first input of the variable length encoder 5 and the input of the inverse quantizer 6.

The output of the inverse quantizer 6 is the inverse DCT converter 7
To the first input of adder 8. The output of the adder 8 is input to the frame memory 9. The output of the frame memory 9 is given to the second input of the second motion vector detector 10c and the first input of the prediction image creating circuit 11. The output of the first motion vector detector 10b is given to the third input of the second motion vector detector 10c. The output of the second motion vector detector 10c is given to the second input of the prediction image creation circuit 11 and the second input of the variable length encoder 5.

The first output of the predictive image forming circuit 11 is given to the second input of the subtracter 2 and the second input of the adder 8. The second output of the prediction image creating circuit 11 is given to the third input of the variable length encoder 5. The output of the variable length encoder 5 is given to the input of the buffer 12a. Buffer 12
The first output of a is transmitted or stored externally, and the second output of the buffer 12a is given to the second input of the valid block determiner 13a.

Next, the operation will be described. First, the frame memory 1a stores an input moving image, rearranges it in the order of encoding, and outputs it from the first output to the subtractor 2. The subtractor 2 obtains the difference between the image output from the frame memory 1a and the predicted image output from the predicted image creation circuit 11. The prediction error output from the subtractor 2 is D
The CT converter 3 divides the image into blocks of 8 pixels × 8 lines, and performs DCT conversion for each block. The effective block determiner 13a uses this DCT coefficient to determine whether to encode each block. For example, if the absolute value of the DC coefficient is less than the threshold value, the input image is determined to be almost the same as the predicted image, and the block is determined to be an invalid block. If the absolute value of the DC coefficient is greater than or equal to the threshold value, it is determined that the block is a valid block that requires encoding.

The quantizer 4a quantizes the transform coefficient output from the DCT transformer 3. At this time, for the blocks determined to be invalid blocks by the valid block determiner 13a, all the transform coefficient values are replaced with 0. The quantized transform coefficient is input to the variable length encoder 5. The variable-length encoder 5 has all quantized transform coefficient values of 0.
If, then the information that the block is an invalid block is encoded; otherwise, the information that the block is an effective block and each transform coefficient are encoded.

On the other hand, the quantized transform coefficient is decoded by the inverse quantizer 6 and the inverse DCT transformer 7. The decoded prediction error is added to the predicted image output from the predicted image creation circuit 11 in the adder 8. The image output from the adder 8 is a decoded image, and the frame memory 9
Is input to

The predicted image is the first motion vector detector 1
0b, the second motion vector detector 10c, and the predicted image generation circuit 11. First, the first motion vector detector 10b detects the motion of the image output from the first output of the frame memory 1a, using the input image stored in the frame memory 1a as a reference image.

For example, the image output from the first output of the frame memory 1a is divided into macroblocks of 16 pixels × 16 lines, and the motion vector of each macroblock is obtained with pixel accuracy. At this time, when the input image is a frame composed of two interlaced fields,
A motion vector for 16 pixels × 16 lines of the frame is obtained, and this macroblock is divided into 16 pixels × 8 lines of two fields to obtain the motion vectors of the two fields. When the image output from the first output of the frame memory 1a is a B picture, the motion vector is calculated for each of the two images stored in the frame memory 1a as reference images.

The second motion vector detector 10c searches for the motion vector with the motion vector output from the first motion vector detector 10b as the center and the decoded image stored in the frame memory 9 as the reference image. I do. For example, centering on the motion vector detected by the first motion vector detector 10b with pixel accuracy, a motion vector is searched from a total of nine motion vectors including this motion vector and the motion vectors at the surrounding half pixel positions.

That is, the prediction distortion is obtained for each of the nine motion vectors, and the vector with the smallest prediction distortion is output as the second motion vector. The prediction distortion at this time is obtained using the output of the frame memory 1a and the decoded image stored in the frame memory 9. When the first motion vector detector 10b outputs a plurality of motion vectors, each motion vector is centered and a search is performed to correct the motion vector.

The predicted image creating circuit 11 creates a predicted image from the decoded image stored in the frame memory 9 based on the motion vector output from the second motion vector detector 10c. For example, when the second motion vector output from the second motion vector detector 10c indicates the position of a half pixel, the pixel value at the position of the half pixel is calculated from the surrounding four pixels. When the input image is a frame composed of two interlaced fields,
A prediction image is created for each of the frame motion vector and the field motion vector, and the one with the smaller prediction distortion is selected and output.

When the image output from the first output of the frame memory 1a is a B picture, two predicted images are created from two reference images, and the average image of these two predicted images is calculated. Created and selecting one with less prediction distortion out of a total of three predicted images. Further, when performing intra-image coding, all pixel values of the predicted image are set to 0 and output. The prediction image creation circuit 11 creates and outputs the prediction image as described above, and also outputs the prediction mode indicating which prediction image is selected.

The prediction mode output from the predictive image creating circuit 11 and the motion vector output from the second motion vector detector 10c are input to the variable length encoder 5 and encoded. The variable-length encoder 5 superimposes these coded data and the coded data of the transform coefficient, and outputs the bitstream to the buffer 12a. Buffer 1
2a outputs this bit stream to the outside and outputs the code amount in the buffer to the valid block determiner 13a. The valid block determiner 13a changes the threshold value based on the code amount in the buffer to control the number of valid blocks to be coded.

In the fifth embodiment, the effective block is determined using the absolute value of the DC coefficient. However, not only the DC coefficient but also the plurality of conversion coefficients are used to obtain the evaluation value indicating the magnitude of the difference signal. Alternatively, the effective block may be determined based on the obtained evaluation value. In particular, when the coding rate is very low, strict restoration of high frequency components is not required, so only the magnitude of the low frequency components of the difference signal is evaluated, so the low frequency coefficient of the transform coefficients is used for evaluation. It may be configured to calculate the value, but when the coding rate is relatively high, it is configured to calculate the evaluation value using all transform coefficients to determine whether the input image and the predicted image are almost the same. You may.

Further, in the fifth embodiment, the DCT transform is used for encoding the differential signal, but the encoding system is not limited to this, and other orthogonal transform, DPCM, vector quantization, subband encoding, etc. Also in the case of using, it is possible to obtain the evaluation value indicating the magnitude of the difference signal and determine the effective block in the same manner.

Furthermore, in the fifth embodiment, the moving picture coding apparatus is configured by combining the first embodiment and the fourth embodiment. However, the combination of the first embodiment and the third embodiment, the second embodiment. And the combination of Embodiment 3 and the combination of Embodiments 2 and 4 can be configured in the same manner, and a moving image encoding apparatus suitable for low rate encoding can be obtained.

[0111]

As described above, the moving picture coding apparatus according to the first invention generates a difference signal between an input image and a predicted image, divides this difference signal into a plurality of pixels and blocks the difference signal. Obtain an evaluation value indicating the magnitude of the low-frequency component of each block, and if this evaluation value is within the threshold range, determine that the block is an invalid block and code only the information indicating that it is an invalid block. If the difference signal is composed of only high-frequency components that are not required to be strictly restored, it is determined that the block is a valid block if it is outside the range of the threshold, and the block is encoded. Since the encoding is not performed, the code amount can be reduced. Further, in the fixed-rate coding, the code amount of the effective block can be increased by the amount of the non-coding, so that the image quality of the decoded image of the effective block is improved.

Since the moving picture coding apparatus according to the second invention controls the threshold value based on the generated code quantity in the moving picture coding apparatus according to the first invention, it is effective according to the target code quantity. The number of blocks determined to be blocks can be controlled.

The moving picture coding apparatus according to the third invention generates a difference signal between an input image and a predicted image, collects the difference signal for each of a plurality of pixels into blocks, and performs orthogonal transformation for each block for transformation. Obtain a coefficient, obtain an evaluation value indicating the magnitude of the low-frequency component of the difference signal from the low-frequency coefficient of this conversion coefficient, and if this evaluation value is within the threshold range, determine that the block is an invalid block Then, only the information indicating that the block is an invalid block is coded, and if it is outside the threshold range, the block is determined to be a valid block and the transform coefficient of the block is coded. When the block is composed of only high-frequency components that are not required to be strictly restored, the block is not coded, so that the code amount can be reduced. Further, in the fixed-rate coding, the code amount of the effective block can be increased by the amount of the non-coding, so that the image quality of the decoded image of the effective block is improved. Furthermore, since the conversion coefficient is used to determine the effective block, it is possible to obtain the evaluation value indicating the magnitude of the low frequency component of the difference signal with a simple configuration.

The moving picture coding apparatus according to the fourth invention divides the input picture into areas consisting of a plurality of pixels, detects a plurality of motion vectors indicating the movement of the moving picture for each area, and based on this motion vector, A plurality of motion-compensated prediction images are created from the decoded images of the already coded images, the prediction distortion of each prediction image is obtained to obtain a prediction image with the minimum prediction distortion, and the difference between this prediction image and the input image A signal is generated, and when the minimum prediction distortion is within the threshold range, it is determined that the area is an invalid block, only the information indicating the invalid block is encoded, and the signal is outside the threshold range. In this case, the region is determined to be a valid block and the difference signal of the region is encoded. Therefore, when the input image and the prediction image are almost the same, the region is not encoded. It can be reduced. Further, in fixed-rate encoding, the code amount of the effective block can be increased by the amount of no encoding, so that the image quality of the decoded image of the effective block is improved. Further, since the prediction distortion used in the selection of the prediction image is used for the determination of the effective block, the circuit for obtaining the magnitude of the difference signal is unnecessary.

In the moving picture coding apparatus according to the fifth aspect of the present invention, the threshold value is controlled based on the generated code quantity in the moving picture coding apparatus according to the fourth invention. Therefore, according to the target code quantity. The number of blocks determined to be valid blocks can be controlled.

The moving picture coding apparatus according to the sixth aspect of the present invention includes a frame memory for storing a plurality of input images, and a plurality of motion vectors indicating the motion of the moving picture are obtained from the plurality of images stored in the frame memory. Since the detection is performed, the motion vector can be detected in the input image with no image quality deterioration, and the motion vector detection accuracy is improved, so that the motion can be faithfully grasped. In addition, a plurality of motion-compensated prediction images are created from the decoded images of the images already encoded based on the plurality of motion vectors, the prediction distortion of each prediction image is obtained, and the prediction image that gives the minimum prediction distortion is obtained. ,
An optimum motion vector can be selected in the sense that the prediction error is reduced and the code amount is reduced in consideration of image quality deterioration of the decoded image.

The moving picture coding apparatus according to the seventh invention comprises a frame memory for storing a plurality of input images, and detects a motion vector indicating the motion of the moving picture from the plurality of images stored in the frame memory. , Since the motion vector is searched centering on this motion vector and the motion vector is corrected, a vector close to the motion vector obtained using the input image with no image quality deterioration can be selected as the motion vector. A close motion vector can be detected. Further, since the motion vector is searched by using the decoded image of the already coded image to correct the motion vector, it is possible to select the motion vector in consideration of the deterioration of the image quality of the decoded image.

The moving picture coding apparatus according to the eighth invention comprises a frame memory for storing a plurality of input images, and a plurality of motion vectors indicating the motion of the moving picture are obtained from the plurality of images stored in the frame memory. Since the detection is performed, the motion vector detection accuracy is improved, and the motion can be grasped faithfully. In addition, a plurality of motion-compensated prediction images are created from the decoded images of the images already encoded based on the plurality of motion vectors, the prediction distortion of each prediction image is obtained, and the prediction image that gives the minimum prediction distortion is obtained. The optimum predicted image can be selected in the sense that the prediction error is reduced in consideration of the image quality deterioration of the decoded image.

Further, a difference signal between the input image and the predicted image is generated, and the difference signal is divided into blocks for each of a plurality of pixels to obtain an evaluation value indicating the magnitude of the difference signal of each block. If it is within the threshold range, it is determined that the block is an invalid block and only the information indicating that it is an invalid block is coded. If it is outside the threshold range, the block is a valid block. Since the determination is made and the block is coded, if the input image and the predicted image are almost the same, the block is not coded, so that the code amount can be reduced. Further, in fixed-rate encoding, the code amount of the effective block can be increased by the amount not encoded, so that the image quality of the decoded image of the effective block is improved. Further, as described above, the optimum predicted image can be selected in the sense of reducing the prediction error, so that the number of blocks determined to be invalid blocks increases and the code amount can be reduced.

The moving picture coding apparatus according to the ninth invention comprises a frame memory for storing a plurality of input images, and detects a motion vector indicating the motion of the moving picture from the plurality of images stored in the frame memory. Since the motion vector is searched centering on this motion vector and the motion vector is corrected, a motion vector close to the actual motion can be detected. Further, since the motion vector is searched for by using the decoded image of the already coded image to correct the motion vector, it is possible to select the motion vector having a small prediction error in consideration of the image quality deterioration of the decoded image.

Further, a difference signal between the input image and a motion-compensated predicted image based on the corrected motion vector is generated, and the difference signal is divided into blocks for each of a plurality of pixels to indicate the magnitude of the difference signal of each block. If the evaluation value is obtained and the evaluation value is within the threshold range, it is determined that the block is an invalid block and only the information indicating the invalid block is encoded. Determines that the block is a valid block and encodes the block. Therefore, when the input image and the prediction image are almost the same, the block is not encoded, so the code amount can be reduced. . Further, in fixed-rate encoding, the code amount of the effective block can be increased by the amount not encoded, so that the image quality of the decoded image of the effective block is improved. Further, as described above, since a motion vector with a small prediction error can be selected, the number of blocks determined to be invalid blocks increases, and the code amount can be reduced.

[Brief description of drawings]

FIG. 1 is a block diagram showing a moving picture coding apparatus according to Embodiment 1 of the present invention.

FIG. 2 is a diagram showing an example of DCT transform coefficients.

FIG. 3 is a block diagram showing a moving picture coding apparatus according to Embodiment 2 of the present invention.

FIG. 4 is a block diagram showing a moving picture coding apparatus according to Embodiment 3 of the present invention.

FIG. 5 is a block diagram showing a moving picture coding apparatus according to Embodiment 4 of the present invention.

FIG. 6 is a block diagram showing a moving picture coding device according to a fifth embodiment of the present invention.

FIG. 7 is a block diagram showing a conventional moving picture encoding device.

FIG. 8 is a conceptual diagram of a motion compensation predictive coding system.

[Explanation of symbols]

1, 1a frame memory, 2 subtractor, 3 DCT converter, 4, 4a, 4b quantizer, 5 variable length encoder, 6
Inverse quantizer, 7 inverse DCT converter, 8 adder, 9
Frame memory 10, 10a, 10b, 10c Motion vector detector, 11, 11a Predicted image generation circuit, 1
2, 12a buffer, 13a, 13b valid block determiner, 14, 15 encoder.

─────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] June 27, 1996

[Procedure amendment 1]

[Document name to be amended] Statement

[Correction target item name] 0047

[Correction method] Change

[Correction contents]

As the low-frequency coefficient, t (n, m) (0
≤ n + m <8),

[Procedure amendment 2]

[Document name to be amended] Statement

[Correction target item name] 0048

[Correction method] Change

[Correction contents]

[0048]

[Equation 2]

[Procedure 3]

[Document name to be amended] Statement

[Correction target item name] 0050

[Correction method] Change

[Correction contents]

Further, the evaluation value d3 may be obtained as d3 = max {| t (n, m) |; 0≤n + m <8}, and the determination may be made based on this evaluation value d3.

Claims (9)

[Claims] 1. A motion vector detection unit that divides an input image into regions composed of a plurality of pixels and detects a motion vector indicating the motion of a moving image for each region, and has already encoded based on the detected motion vector. A prediction image generation unit that generates a motion-compensated prediction image from a decoded image of the image; a subtractor that generates a difference signal between the input image and the prediction image;
This difference signal is divided into blocks for each of a plurality of pixels to obtain an evaluation value indicating the magnitude of the low-frequency component of each block, and if this evaluation value is within the threshold range, it is determined that the block is an invalid block. Also, a moving image coding apparatus including a valid block determination unit that determines that the block is a valid block when the block is outside the threshold range, and a coder that encodes the block determined to be the valid block. 2. The moving picture coding apparatus according to claim 1, further comprising means for controlling the threshold value based on a code amount output from the encoder. 3. A motion vector detection unit that divides an input image into regions composed of a plurality of pixels and detects a motion vector indicating the motion of the moving image for each region, and has already encoded based on the detected motion vector. A prediction image generation unit that generates a motion-compensated prediction image from a decoded image of the image; a subtractor that generates a difference signal between the input image and the prediction image;
This difference signal is grouped into a plurality of pixels to form a block, and an orthogonal transformation circuit that performs orthogonal transformation for each block, and the magnitude of the low-frequency component of the difference signal from the low-frequency coefficient among the transformation coefficients output from this orthogonal transformation circuit If the evaluation value is within the threshold range, it is determined that the block is an invalid block, and if the evaluation value is outside the threshold range, the block is determined to be a valid block. And a coder for coding the transform coefficient output from the orthogonal transform circuit for the block determined to be the valid block. 4. A motion vector detection unit that divides an input image into regions composed of a plurality of pixels and detects a plurality of motion vectors indicating the motion of a moving image for each region, and a code that has already been coded based on the detected motion vectors. A prediction image creating unit that creates a plurality of motion-compensated prediction images from the decoded images of the converted images to obtain the prediction distortion of each prediction image, and outputs the prediction image that gives the minimum prediction distortion, the input image and the prediction image And a subtractor that generates a difference signal between the region and the minimum prediction distortion is within the range of the threshold, it is determined that the region is an invalid block, and if it is outside the range of the threshold, the region is a valid block. A moving picture coding apparatus, comprising: a valid block determining unit that determines that the valid block is determined to be a valid block; 5. The moving picture coding apparatus according to claim 4, further comprising means for controlling the threshold value based on a code amount output from the encoder. 6. A frame memory for storing a plurality of input images, a motion vector detection unit for detecting a plurality of motion vectors indicating a motion of a moving image from the plurality of images stored in the frame memory, and the detected motion vector detecting unit. Based on the motion vector, a plurality of motion-compensated prediction images are created from the decoded images of the images already encoded, and the prediction distortion of each prediction image is obtained.
A moving image coding apparatus comprising: a predicted image generating unit that outputs a predicted image that gives a minimum predicted distortion; and a subtractor that generates a difference signal between the input image and the predicted image. 7. A frame memory for storing a plurality of input images, a first motion vector detection unit for detecting a motion vector indicating a motion of the moving image from the plurality of images stored in the frame memory, and a first motion vector detection unit for detecting the motion vector. A second motion vector detection unit that performs a motion vector search using a decoded image of an already coded image centering on the motion vector output from the one motion vector detection unit, and this second motion vector detection unit A prediction image generation unit that generates a motion-compensated prediction image from a decoded image of an image that has already been encoded based on the motion vector output from the subtraction unit that generates a difference signal between the prediction image and the input image. Video coding device. 8. A frame memory for storing a plurality of input images, a motion vector detection unit for detecting a plurality of motion vectors indicating the motion of a moving image from the plurality of images stored in the frame memory, and the detected motion vector. Based on the motion vector, a plurality of motion-compensated prediction images are created from the decoded images of the images already encoded, and the prediction distortion of each prediction image is obtained.
A prediction image generation unit that outputs a prediction image that gives a minimum prediction distortion, a subtractor that generates a difference signal between the input image and the prediction image, and a difference between each block by dividing the difference signal into a plurality of pixels. The evaluation value indicating the magnitude of the signal is obtained, and when this evaluation value is within the threshold range, it is determined that the block is an invalid block, and when it is outside the threshold range, the block is a valid block. A moving picture coding apparatus, comprising: a valid block determining unit that determines that there is a block; and a coder that encodes a block that is determined to be the valid block. 9. A frame memory for storing a plurality of input images, a first motion vector detection unit for detecting a motion vector indicating a motion of the moving image from the plurality of images stored in the frame memory, and a first motion vector detection unit for detecting the motion vector. A second motion vector detection unit that performs a motion vector search using a decoded image of an already coded image centering on the motion vector output from the one motion vector detection unit, and this second motion vector detection unit A prediction image generation unit that generates a motion-compensated prediction image from a decoded image of an image that has already been encoded based on the motion vector output from the subtraction unit that generates a difference signal between the input image and the prediction image; The difference signal is divided into blocks for each of a plurality of pixels, and an evaluation value indicating the magnitude of the difference signal of each block is obtained. If the evaluation value is within the threshold range, the block does not exist. A valid block determining unit that determines that the block is a valid block, and determines that the block is a valid block when the block is outside the threshold range, and an encoder that encodes the block that is determined to be the valid block. A moving picture coding device provided.