JPH07298270A - Inter-motion compensation frame prediction coder - Google Patents

Abstract

PURPOSE:To provide an inter-motion compensation frame prediction coder by which a motion vector is subjected to compression coding efficiently and an image with excellent image quality is reproduced. CONSTITUTION:The inter-motion compensation frame prediction coder applying prediction coding to a motion image signal by using a picture element at a position of a reference frame by a motion vector as a reference picture element is provided with an object extract means 12 extracting an area in an image of each object in an input frame, a motion vector detection means 13 detecting a motion vector for each extracted object, an acceleration detection means 14 detecting an acceleration vector of each object based on a motion vector of the same object detected in a preceding frame and the detected motion vector, and a multiplexer means 16 multiplexing the detected acceleration vector with information subjected to prediction coding to apply variable length coding to the multiplexed information. The information is compressed by converting the motion vector into the acceleration vector being the difference between motion vectors, high efficiency coding is attained and lots of code quantity is allocated to a quantization coefficient giving much effect onto the image quality to improve the image quality of the reproduced image.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a motion-compensated interframe predictive coding apparatus used for signal processing of video conferences, videophones, cable televisions, etc., and particularly, enables efficient efficient compression coding. is there.

[0002]

2. Description of the Related Art In recent years, digital signal processing technology has been developed,
It has become possible to compress and encode moving images for transmission, and video conferences and video phones have come into practical use.

In coding a moving image, in order to improve coding efficiency, the pixel value of a pixel to be coded next is predicted using the transmitted pixel information, and the predicted value is used instead of the actual pixel value. Predictive coding is performed to code the error of.
Among them, the motion-compensated interframe predictive coding method detects the amount of movement of the subject between consecutive frames and uses the corresponding pixel of the previous frame when the screen position is translated by that amount of movement for prediction. The prediction accuracy is significantly improved, and a high compression rate can be obtained.

Further, in the coding of image signals, orthogonal transform coding is also performed in which the image signals are subjected to discrete cosine transform (DCT) or the like to transform them into frequency domain signals. By this conversion, highly correlated image signals are concentrated in the low frequency components of the spectrum. Although this conversion alone does not result in data compression, many bits are assigned to low-frequency components, and high-frequency components with low power are encoded with a small number of bits to reduce the required number of bits on average. be able to.

As shown in FIG. 3, a conventional motion compensation interframe predictive coding apparatus which implements a combination of each of the above-described systems temporarily outputs the input image and a terminal 1 to which a digitized image (input image) is input. A frame memory 2 that outputs the data in block units after storing, a subtracter 3 that outputs a difference (prediction error) between an input image and its prediction image, and a prediction error DC
An orthogonal transform unit 4 that transforms into a frequency domain by a transform method such as T (discrete cosine transform) and outputs an orthogonal transform coefficient that represents power for each frequency; a quantizer 5 that quantizes the orthogonal transform coefficient; The inverse quantization unit 8 that inversely quantizes the converted orthogonal transform coefficient, the inverse orthogonal transform unit 9 that obtains a prediction error image by performing the inverse orthogonal transform of the input orthogonal transform coefficient, and the prediction that is input from the inverse orthogonal transform unit 9. The adder 10 that outputs the reproduced image by adding the error image and the predicted image used in the subtractor 3, the frame memory 11 that stores the reproduced image, and the image obtained from the frame memory 11 in the area for each object In which direction is the region detected by the region dividing unit 17 between the region dividing unit 17 for dividing and labeling and the reproduced image of the previous frame input from the frame memory 11 and the input image input from the frame memory 2. How much you moved A motion vector detecting section 13 for detecting a motion vector Wath, code multiplexing unit 16 for multiplexing a label or the like of motion vectors and area output from the quantized data and motion vector detection section 13 output from the quantization unit 5
And a variable length coding unit 6 for variable length coding the multiplexed code, and an output terminal 7 for outputting the coded moving image data.

In this apparatus, a digitized input image to be compressed and encoded is input from the terminal 1 and is temporarily stored in the frame memory 2. The frame memory 2 outputs the input image stored in a predetermined block unit to the subtractor 3, and the motion vector detection unit 13 outputs the input image.
It outputs the pixel data required to retrieve the motion vector.

An input image from the frame memory 2 and a predicted image for the input image are input to the subtractor 3 from the frame memory 11, and the subtractor 3 inputs the difference, the prediction error image, to the orthogonal transform unit 4. Output. The orthogonal transformation unit 4
The input prediction error is converted into a frequency domain by a conversion method such as DCT, and an orthogonal transform coefficient representing the power of each frequency is output. The quantization unit 5 quantizes the input orthogonal transformation coefficient with the quantization parameter given by the variable length coding unit 6, and the quantization coefficient and the quantization parameter at that time are quantized to the code multiplexing unit 16 and the dequantization unit 8. Output.

The dequantization unit 8 dequantizes the input quantized coefficient with the quantization parameter used at the time of quantization, and outputs it to the inverse orthogonal transform unit 9. The inverse orthogonal transform unit 9 performs an inverse orthogonal transform on the input orthogonal transform coefficient to obtain a prediction error image, and the adder 10
Output to. The adder 10 adds the prediction error image input from the inverse orthogonal transform unit 9 and the prediction image used in the subtractor 3 to reproduce the input image, and outputs it to the frame memory 11. This reproduced image is stored in the frame memory 11 for use in region division of the next frame, motion vector detection, and prediction image creation.

The area dividing unit 17 extracts, for example, edge information of an object from the image obtained from the frame memory 11 and divides the area for each object. Then, a label is attached to each area according to a predetermined procedure, and the result is output to the motion vector detecting unit 13. The motion vector detection unit 13 includes an input image input from the frame memory 2 and the frame memory 11
The motion vector of each area detected by the area dividing unit 17 is detected using the reproduced image of the previous frame input from the.
The detected motion vector is output to the code multiplexing unit 16 and the frame memory 11 together with the label of the corresponding area.

The frame memory 11 is a motion vector detecting section.
A predicted image is read out based on the motion vector input from 13, and this is output to the subtractor 3 and the adder 10.

The code multiplexing unit 16 includes a quantization coefficient and a quantization parameter output from the quantization unit 5, a motion vector output from the motion vector detection unit 16 and a label of a region corresponding to the motion vector, and other decoding. The parameters required for encoding are multiplexed and output to the variable length coding unit 6.

The variable length coding unit 6 performs variable length coding on the multiplexed code and outputs it to the outside through the output terminal 7. At this time, in order to compress the code amount of the motion vector, the difference between the motion vector and the motion vector of the adjacent area is also variable-length coded.

In addition, in the coding system using the variable length code, the amount of generated bits changes with time, so that the buffer supplying data to the transmission line for transmitting data at a fixed rate overflows or underflows. There is a risk of doing it. In order to prevent such a situation, the variable length coding unit 6 outputs the quantization parameter to the quantization unit 5 and controls the code amount of the compression code sent from the output terminal 7. That is, when the generated code amount is larger than the target amount, the quantization parameter for expanding the quantization width is set to the quantization unit 5.
And the amount of code generated by the quantizer 5 is adjusted by feeding back a quantization parameter that narrows the quantization width when the amount is small.

In the conventional motion-compensated interframe predictive coding apparatus, the moving image is compression-coded and transmitted in this way.

[0015]

However, in the coding by the conventional motion compensation interframe predictive coding apparatus, when the number of regions of the image divided by the region dividing unit 17 is large, the code amount of the motion vector is As a result, the number of codes assigned to the quantized coefficients that affect the image quality decreases, and as a result, the image quality of the reproduced image deteriorates.

The present invention solves the above-mentioned conventional problems. Even when the number of regions into which an image is divided is large, the motion vector can be efficiently compression-coded, thereby providing excellent image quality. It is an object of the present invention to provide a motion-compensated inter-frame predictive coding device that enables image reproduction of the above.

[0017]

Therefore, according to the present invention, in a motion-compensated interframe predictive coding apparatus for predictive-coding a moving picture signal using a position shifted by a motion vector in a reference frame as a reference pixel, an input frame Object extracting means for extracting the region in the image of each object inside, motion vector detecting means for detecting the motion vector of each extracted object, the detected motion vector and the motion of the same object detected in the previous frame An acceleration detecting means for detecting the acceleration vector of each object based on the vector and a multiplexing means for multiplexing the detected acceleration vector with the information coded by prediction are provided, and the multiplexed information is variable length. It is configured to be encoded.

Further, object integrating means for integrating adjacent objects having the same motion vector among the objects extracted by the object extracting means is provided, and the acceleration vector for the integrated object is narrowed down to one, and the multiplexing means is provided. It is configured to be sent to.

[0019]

As described above, by converting the motion vector into the acceleration vector which is the difference between the motion vectors, the information is compressed and high efficiency coding becomes possible. Therefore, a large amount of code can be assigned to the quantization coefficient that has a great influence on the image quality, and the image quality of the reproduced image can be improved.

Further, by integrating the objects having the same motion, it becomes possible to further compress the information regarding the motion vector.

[0021]

BEST MODE FOR CARRYING OUT THE INVENTION As shown in FIG. 1, a motion compensation interframe predictive coding apparatus according to an embodiment of the present invention includes an object extracting unit 12 for extracting each object from an image based on edge information of the object, and the like.
The acceleration detection unit 14 that outputs the difference between the motion vector of the previous frame of each object and the motion vector of the current frame as an acceleration vector, and the adjacent objects that have the same motion vector are combined into one object. An object integration unit 15 that outputs an acceleration vector for each object to the code multiplexing unit 16. Other configurations are the same as those of the conventional device (FIG. 3).

As shown in FIG. 2, the acceleration detecting section 14 has an input terminal 21 for inputting a motion vector and label information.
And the label conversion unit 22 that outputs the label of the corresponding (same) object in the previous frame when the label information is input, and the motion vector value of the corresponding previous frame when the label is output from the label conversion unit 22. Output memory 23
And a subtractor 24 for calculating the difference between the motion vector of the previous frame and the motion vector of the current frame, and an output terminal 25 for outputting the difference calculated by the subtractor 24 as an acceleration vector.

The input terminal 1 of this motion compensation inter-frame predictive coding apparatus receives a digitized input image which is compression-coded from now on, is temporarily stored in the frame memory 2, and is then stored in the subtracter 3 in advance. Pixel data necessary for searching for a motion vector is output to the motion vector detection unit 13 in the determined block unit.

The subtracter 3 outputs the difference between the input image input from the frame memory 2 and the predicted image input from the frame memory 11 to the orthogonal transformation unit 4 as a prediction error, and the orthogonal transformation unit 4 Is the prediction error D
The quantization unit 5 transforms the signal into a frequency domain by a transformation method such as CT and outputs an orthogonal transformation coefficient representing the power of each frequency.
The orthogonal transform coefficient is quantized by the quantization parameter given by the variable length coding unit 6, and the quantized coefficient and the quantization parameter at that time are output to the code multiplexing unit 16 and the dequantization unit 8.

The inverse quantization unit 8 inversely quantizes the quantized coefficient with the quantization parameter used at the time of quantization, and the inverse orthogonal transform unit 9
Is an inverse orthogonal transform of the orthogonal transform coefficient obtained by inverse quantization,
The adder 10 reproduces the input image by adding the prediction error image obtained by the inverse orthogonal transform and the prediction image used by the subtractor 3, and the frame memory 11 divides this reproduced image into the regions of the next frame. , For use in motion vector detection and prediction image generation.

When the object extracting unit 12 obtains the image from the frame memory 11, the object extracting unit 12 extracts each object based on, for example, edge information of the object, and labels each object according to a predetermined procedure.
The area of the object and the label are output to the motion vector detection unit 13 as label information. The motion vector detection unit 13 is configured to detect the input image input from the frame memory 2 and the frame memory.
The motion vector of each object extracted by the object extraction unit 12 is detected from the reproduced image of the previous frame input from 11. The detected motion vector is input to each of the acceleration detection unit 14, the object integration unit 15, and the frame memory 11 together with the label information of the corresponding object.

The frame memory 11 creates a predicted image from the already stored reproduced image of the previous frame based on the detected motion vector, and outputs it to the subtractor 3 and the adder 10.

The acceleration detecting section 14 includes a motion vector detecting section.
When the motion vector detected by 13 and the label information of the object detected by the object extraction unit 12 are input through the terminal 21, the label conversion unit 22 that has received the label information of the object labels the corresponding (same) object label in the previous frame. Is output to the memory 23. When the label information is output from the label conversion unit 22, the memory 23 that stores the motion vector of the previous frame and the label information reads the motion vector value of the previous frame corresponding to the label and outputs it to the subtractor 24. To do.

The subtractor 24 calculates the difference between the motion vector of the previous frame output from the memory 23 and the motion vector of the current frame input from the terminal 21, and uses this value as the acceleration vector through the terminal 25 to obtain the object integration unit. Output to 15.
Further, the memory 23 stores the label information and the motion vector value input from the terminal 21 in order to calculate the acceleration vector of the next frame.

The object integrating unit 15 integrates adjacent objects having the same motion vector as one object based on the motion vector of each object input from the motion vector detecting unit 13,
The acceleration vector and label information for each divided object after this integration processing are output to the code multiplexing unit 16. For a plurality of objects that have been integrated into one, integrated information indicating that they have been integrated and an acceleration vector for one of those objects are output to the code multiplexing unit 16, and accelerations for the remaining objects are output. The vector is not sent to the code multiplexing unit 16.

The code multiplexing unit 16 includes the quantization coefficient and the quantization parameter output from the quantization unit 5, the acceleration vector output from the object integration unit 15, and the label information and integration information of the object corresponding to the acceleration vector. , And other parameters necessary for decoding are multiplexed and output to the variable length coding unit 6, and the variable length coding unit 6 performs variable length coding on these multiplexed information and outputs them to the outside through the output terminal 7. To do.
At the same time, the variable length coding unit 6 outputs the quantization parameter to the quantization unit 5 to control the code amount of the compression code sent from the output terminal 7.

As described above, in the motion compensation interframe predictive coding apparatus of the embodiment, the acceleration vector is calculated on the basis of the motion vector and is coded. This encoding of the acceleration vector makes it possible to compress information as compared with the case of directly encoding the motion vector. This is because an ordinary object is less likely to receive a force from the outside, and in the case of an object that does not receive a force from the outside, its motion becomes a uniform linear motion, and the acceleration, which is a change in the velocity vector. Is 0. Further, even when the motion of the object has acceleration, the acceleration vector is defined as the amount of change in the motion vector due to the time difference between the frames. Therefore, the short time between these two frames (in the case of the NTSC signal, this time is 1 This is because the velocity change of the object that occurs in / 30 seconds) is very small.

Further, in the apparatus of the embodiment, the object integrating unit 15 integrates adjacent objects having the same motion vector into one, and the code amount can be reduced through this integration. Since the object extraction unit 12 detects an object from image information such as edge information, even if it is actually one object, it may be detected as two different objects. They are integrated into one by the integration unit 15. Further, when the individual objects themselves are stationary, but the objects in the image move due to an operation such as panning of the camera, those objects are integrated and the code amount is compressed. In addition, when a plurality of adjacent objects are integrated, on the receiving side, the motion vector for that one object is
By obtaining it based on the acceleration vector, the motion vector of another object can be known.

As described above, since the motion compensation interframe predictive coding apparatus of the embodiment can compress the information about the motion vector, it is possible to allocate a large amount of code to the quantization coefficient in the variable length coding. As a result, a high quality reproduced image can be transmitted.

[0035]

As is clear from the above description of the embodiments, the motion-compensated interframe predictive coding apparatus of the present invention calculates the acceleration vector for each object and codes the acceleration vector. The amount of information can be efficiently compression-coded.

Further, by integrating the objects having the same motion vector, the amount of the acceleration vector to be coded can be reduced and the compression rate of the information about the motion vector can be further increased.

As a result, it is possible to increase the code amount assigned to the quantization coefficient that greatly affects the image quality, and it is possible to transmit a reproduced image of high quality.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a motion compensation interframe predictive coding apparatus according to the present invention,

FIG. 2 is a block diagram showing the configuration of an acceleration detection unit in the apparatus of the embodiment,

FIG. 3 is a block diagram showing a configuration of a conventional motion compensation interframe predictive coding device.

[Explanation of symbols]

 1 Input Terminal 2, 11 Frame Memory 3, 24 Subtractor 4 Orthogonal Transformation Section 5 Quantization Section 6 Variable Length Coding Section 7 Output Terminal 8 Inverse Quantization Section 9 Inverse Orthogonal Transformation Section 10 Adder 12 Object Extraction Section 13 Motion Vector Detection unit 14 Acceleration detection unit 15 Object integration unit 16 Code multiplexing unit 17 Region division unit 21 Input terminal 22 Label conversion unit 23 Memory 25 Output terminal

Claims (2)

[Claims] 1. A motion-compensated inter-frame predictive coding apparatus for predictive-coding a moving picture signal by using a position shifted by a motion vector in a reference frame as a reference pixel, and determining a region in an image of each object in an input frame. An object extracting means for extracting, a motion vector detecting means for detecting a motion vector of each extracted object, and an acceleration vector of each object based on the detected motion vector and the motion vector of the same object detected in the previous frame. Is provided, and a multiplexing means for multiplexing the detected acceleration vector with the predictively coded information is provided, and the multiplexed information is variable-length coded. Motion compensation interframe predictive coding device. 2. The object extracted by the object extracting means,
2. The motion compensation according to claim 1, further comprising object integrating means for integrating adjacent objects having the same motion vector, and limiting the acceleration vector of the integrated object to one and sending it to the multiplexing means. Interframe predictive coding device.