JPH04178089A - Movement compensation inter-frame coder - Google Patents

Abstract

PURPOSE:To select a quantization width suitable for a line rate by controlling quantizing width of a quantizer not only from information of preceding coding but also information quantity generated in a preceding block. CONSTITUTION:A digital picture data is inputted from a terminal 31 and a prediction value outputted from a frame memory 43 is subtracted by a subtractor 32 and a prediction error being a difference is inputted to an orthogonal transmission circuit 33 and a quantization circuit 34, and a coding circuit 37, in which the data is coded and the relation among a code quantity INF, a quantization value Q and its coefficient B is given by equation (I). A coefficient calculation circuit 36 uses the coefficient B of a block of a preceding same location and a quantization width calculation circuit 35 calculates the quantization width Q giving a generated code quantity INF in matching with the line speed. The transformation coefficient quantized by the circuit 34 is restored into a transformation coefficient by an inverse quantization circuit 40 and inputted to an inverse orthogonal transformation circuit 41, in which the coefficient is restored to the prediction error. An output of an adder 42 is fed to a frame memory 43, in which the data is used for detecting a prediction value of a succeeding frame and a moving vector. The moving vector is detected from the output of the memory 43 and the input from the terminal 31.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a motion compensated interframe coding device that is used in video conferences, video telephones, television monitoring systems, etc. and is adapted to obtain an optimal quantization width.

BACKGROUND OF THE INVENTION FIG. 3 shows the configuration of a conventional motion compensated interframe coding device. In FIG. 3, the digitized image is inputted from a terminal 11 and inputted to one input terminal of a subtracter 12. The predicted value output from the frame memory 20 is also input to the subtraction input terminal of the subtracter 12, and a prediction error is output. This prediction error is input to the orthogonal transform circuit 13 and orthogonally transformed. The transform coefficients are input to the quantization circuit 14 and quantized using a quantization width determined by the amount of remaining buffer in the line buffer 16. The quantized transform coefficients are input to the inverse quantization circuit 22 and the encoding circuit 15. The encoding circuit 15 encodes the transform coefficients quantized by the quantization circuit 14 and the motion vectors detected by the motion vector detection circuit 21, multiplexes them, outputs them to the line buffer 16, and outputs them to the line through the terminal 17. On the other hand, the inverse quantization circuit 22
returns the quantized transform coefficients to the original transform coefficients, performs inverse orthogonal transform in an inverse orthogonal transform circuit 18, and outputs a prediction error.

The prediction error obtained by the inverse orthogonal transform is given to an adder 19. The adder 19 adds the predicted value from the frame memory 20 and the prediction error from the inverse orthogonal transform circuit 18, and outputs a reproduced pixel value to the frame memory 20. The frame memory 20 creates a reproduced image from the reproduced pixel values from the adder 19 and outputs it as a predicted value for the next frame.

Further, the motion vector detection circuit 21 determines a motion vector from the input image input from the terminal 11 and the reproduced image from the frame memory 20 and outputs it.

In this way, the conventional motion compensated interframe coding device described above can also determine the quantization width and output a motion vector.

However, in the conventional motion compensated interframe coding apparatus described above, the quantization width of the quantization circuit 14 is determined from the amount of codes remaining in the line buffer 16, that is, the amount of codes generated in the past. However, there was a problem in that it was difficult to improve the image quality of a certain part of the frame (generate a large amount of code) and control the amount of code for the entire frame.

The present invention solves these conventional problems, and aims to provide an excellent motion compensated interframe code device that can obtain an optimal quantization width.

Means for Solving the Problems In order to achieve the above object, the invention according to claim 1 provides a code generation method based on the generated code amount and quantization width of a block of a past frame at the same position as a block of a frame to be encoded. A coefficient output unit that outputs coefficients for calculating the relationship between code amount and quantization width, and the coefficients output from this coefficient output unit are used to determine the quantization width from a given target amount of generated code. The quantization width calculation unit is also provided with a quantization width calculation unit.

Further, the invention according to claim 2 provides the encoding device according to claim I, further comprising: a motion block count unit that predicts the number of motion-compensated blocks in the current frame from the number of motion-compensated blocks in the past frame; A code amount setting unit that calculates a code amount to be allocated to each block from the number of blocks.

Therefore, according to the invention described in claim 1, the generated code amount of the block for encoding the current frame is reduced by utilizing the property that the correlation between the blocks is high between the blocks at the same position as the previous frame and the current frame. By predicting the amount of code generated for blocks at the same position in the past that have been previously encoded and the quantization width at that time, it is possible to select the quantization width that generates the amount of code that matches the line rate. It has the effect of being able to generate a code amount that matches the line rate.

Further, according to the second aspect of the invention, the number of motion-compensated blocks is predicted from the number of blocks in one frame that has been motion-compensated in the past, and the target value of the code amount is different for each block depending on whether or not motion compensation is performed. By providing this, it is possible to improve the image quality of moving areas or still areas without keeping the frame rate constant.

Embodiment FIG. 1 shows the configuration of an embodiment of the present invention. In FIG. 1, a digitally converted input image is input from a terminal 31, a subtracter 32, and a motion vector detection circuit 44 (motion vector detection section). The subtracter 32 subtracts the predicted value outputted from the frame memory 43 (predicted signal generation section) from the input value (image) inputted from the terminal 31,
The prediction error is output to the orthogonal transform circuit 33. The orthogonal transform circuit 33 orthogonally transforms the prediction error input from the subtracter 32 and outputs transform coefficients to the quantizer circuit 34 (quantizer). The orthogonal transformation performed here is a discrete cosine transformation or the like. The quantization circuit 34 quantizes the transform coefficients input from the orthogonal transform circuit 33 according to the quantization width given by the quantization width calculation circuit 35 (quantization width calculation section). The quantized transform coefficients are output to an encoding circuit 37 (encoding section) and an inverse quantization circuit 40. The encoding circuit 37 encodes the transform coefficients quantized by the quantization circuit 34 and the motion vector detected by the motion vector detection circuit 44, and outputs the encoded signals to the line buffer 38. The line buffer 38 temporarily stores the code encoded by the encoding circuit 37 and outputs it to the line through a terminal 39 at a speed matching the line speed. Furthermore, the encoding circuit 37 calculates the code amount for each block and outputs it to the coefficient calculation circuit 36 along with the quantization width. The coefficient calculation circuit 36 calculates the coefficient B in equation (1) representing the relationship between the quantization width and the amount of generated code that has been determined experimentally, and outputs it to the quantization width calculation circuit 35.

log INF = a X log Q + B
(1) Here, INF is the amount of code generated for each macroblock, a is a fixed value, and B is a coefficient that varies depending on the image.

The quantization width calculation circuit 35 stores the coefficient B output from the coefficient calculation circuit 36, and calculates the line speed by using the coefficient B of the coefficient block of the past block at the same position as the block to be quantized by the quantization circuit 34. The quantization width for generating the amount of generated code corresponding to the amount is calculated and outputted to the quantization circuit 34. Also,
The transform coefficients quantized by the quantization circuit 34 are also input to an inverse quantization circuit 40, where they are returned to transform coefficients, and input to an inverse orthogonal transform circuit 41, where they are returned to prediction errors. Here, since the inverse orthogonal transform circuit 41 performs inverse orthogonal transform on the once quantized transform coefficient, the inversely transformed prediction error includes a quantization error. Therefore, the prediction error of the output of the adder 42 is a different value. The prediction error output from the inverse orthogonal transform circuit 41 is added to the predicted value from the frame memory 43 by an adder 42 to obtain reproduced pixels.

The reproduced pixels output from the adder 42 are stored in the frame memory 4.
3, a reproduced image is created, and used to detect the predicted value and motion vector of the next frame. The motion vector detection circuit 44 detects a motion vector from the reproduced image outputted from the frame memory 43 and the human input image inputted from the terminal 31, and outputs the detected motion vector. The detected motion vector is input to the encoding circuit 37 and the frame memory 43, and used for encoding and outputting predicted values, respectively.

Further, another embodiment is shown in FIG. In addition to the above embodiments, a motion block number prediction circuit 61 and a code amount setting circuit 62 are added. Since the other parts operate in the same manner as in the embodiment shown in FIG. 1, only the different parts will be explained.

Motion block number prediction circuit 61 (motion block prediction unit)
counts the number of blocks in which a motion vector is detected in one frame based on the motion vector information output from the motion vector detection circuit 44, predicts the number of blocks in which a motion vector is detected in the next frame, It is output to the code amount setting circuit 62 (code amount setting section). Here, as a prediction method, it is possible to simply use the number of blocks in which the motion vector of the previous frame was detected or linear prediction from several past frames. The code amount setting circuit 62 determines blocks in the moving area from the line speed, the frame rate of the image to be encoded, the ratio of the code amount of the moving area to the still area, and the number of motion vector detection blocks output from the moving block number prediction circuit 61. A code amount is assigned to each block in the still area, and the amount is applied to the quantization width calculation circuit 35.

The code amount allocation for each block is given by the following equation (2+, (31).

S inf = Tr/ (NB-Mm+R-Mm)
Fr ・= (2) M inf = R−Tr/ (
NB-h+R-Mm) Fr= (3) Here, S i
nf is the code amount per 1 pro/8 in the stationary area, M in
f is the amount of code per block in the motion area, NB is the total number of blocks in one frame, Fr is the frame rate) (Hz
), Tr is the line speed (bps), Mm is the number of blocks in which motion vectors are detected, and R is the ratio of the amount of code between the motion area and the still area. Therefore, by changing the ratio R, it is possible to control the image quality of the encoded image while keeping the frame rate of the encoded frame constant.

That is, by increasing the ratio R, the amount of code allocated to the moving area increases, so the image quality of the moving area can be improved compared to that of the still area.

In this way, according to the above embodiment, the generated code amount is calculated by predicting the relationship between the generated code amount and the quantization width of the corresponding block from the generated code amount of the block at the same position in the past frame and the quantization width. This has the effect that the quantization width can be controlled with higher precision.

In addition, by predicting the number of blocks in the motion area and setting the code amount individually according to the characteristics of the blocks, it is possible to easily control the image quality while maintaining the code amount that matches the line speed for each frame. have

Effects of the Invention As is clear from the above embodiments, the present invention has the following effects.

Quantization is performed not only from past encoding information (amount of code remaining in the buffer) but also from the amount of information generated in past blocks by taking advantage of the property that the correlation with blocks at the same position in past frames is large. Since the quantization width of the quantizer is controlled, it is possible to select the quantization width of the quantizer that generates a code amount more suitable for the line rate.

It also has the advantage that different amounts of code can be given to the moving area and the still area, and the image quality can be controlled for each block while keeping the amount of generated code for the entire frame almost constant.

[Brief explanation of drawings]

FIG. 1 is a schematic block diagram of a motion-compensated interframe coding device in one embodiment of the present invention, FIG. 2 is a schematic block diagram of a motion-compensated interframe coding device in another embodiment of the invention, and FIG. 1 is a block diagram showing the configuration of a conventional motion compensated interframe coding device. 34... Quantization circuit, 35... Quantization width calculation circuit,
37... Encoding circuit, 43... Frame memory, 4
4... Motion vector detection circuit, 61... Motion block number prediction circuit, 62... Code amount setting circuit.

Claims (2)

[Claims] (1) A motion vector detection unit that divides one frame of an input television signal into multiple blocks and detects a motion vector representing image motion for each divided block, and motion compensation based on this motion vector. a prediction signal generation section that generates a prediction signal for the input signal by performing the following steps; a quantization section that quantizes a prediction error signal that is the difference between the prediction signal and the input signal; and a quantization section that quantizes the quantized signal and the motion vector. and a coefficient that outputs a coefficient that changes depending on the characteristics of the image when calculating the quantization width from the past quantization width of the quantization unit and the amount of code generated in the encoding unit. A motion compensated interframe coding device comprising: an output section; and a quantization width calculation section that calculates a quantization width for generating a specified amount of generated codes from the coefficients. (2) A motion block prediction unit that obtains a motion vector from a motion vector detection circuit and predicts the number of blocks to be motion compensated for in the current frame based on the number of blocks for which motion compensation was performed and blocks for which motion compensation was not performed in past encoded frames. and a code amount setting unit that individually allocates code amounts to the blocks subjected to motion compensation by the motion block prediction unit and the blocks not subjected to motion compensation, and supplies them to the quantization width calculation unit. Motion compensated interframe coding device.