JPH02250581A - Highly efficient encoding method for video signal - Google Patents

Abstract

PURPOSE:To highly efficiently encode a video signal by limiting orthogonal transformation coefficients for quantization within a small range. CONSTITUTION:A picture is divided in units of a block consisting of longitudinal eight picture elements X lateral eight picture elements. The picture elements in one block accumulated in a buffer memory 1 are prediction-encoded in the prediction direction fixed beforehand with the use of a DPCM encoder 2. A differential signal is inputted to an orthogonal transformation unit 3, orthogonally transformed on a two-dimensional basis, and the transformed coefficients are fed to a quantized range setting device 4, and it is limited to the quantization range set at every block. Consequently the orthogonally transformed coefficients can be made much concentrated in a small range than usual. Thus the video signal can be highly efficiently encoded.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of the Invention The present invention relates to a high-efficiency encoding method for video signals in which encoding by orthogonal transformation is efficiently performed.

One of the conventional high-efficiency encoding methods for video signals is based on orthogonal transformation.

This divides the screen into small blocks, orthogonally transforms each block, and encodes the coefficients. Generally, in a video signal, there is a strong correlation between pixels, and when orthogonal transformation is applied to a signal divided into small blocks, as shown in Figure 7, coefficients with large values are represented in the low frequency part, that is, in a two-dimensional matrix. It tends to be concentrated in the upper left part of the screen. Furthermore, when a characteristic frequency component is found within a block, the coefficient of the portion corresponding to that frequency appears large.

By limiting the range of quantization to only the vicinity of concentrated coefficients or by making the quantization width variable, the number of bits required for encoding coefficients can be reduced.

A method of setting the quantization range of coefficients in a conventional orthogonal transform is shown in FIGS. 8 to 10. In FIG. 8, the range of quantization is limited to a predetermined area (here, a rectangular area indicated by a broken line). By regarding coefficients outside this range as zero, the number of coefficients to be quantized can be limited.

In FIG. 9, the range of quantization is limited to the smallest rectangular part surrounded by rows and columns in which the DC component (upper left corner) of the coefficient is one vertex and the coefficients are all zero,
Send an address indicating the vertical and horizontal range. Coefficients outside this range are zero and are therefore not sent to the quantizer.

In FIG. 10, the last non-zero coefficient is found by zigzag scanning, and up to this part is quantized. In this case, a method is generally used in which the number of coefficients that are zero and the value of non-zero coefficients are quantized.

In addition to these methods, if it is desired to further reduce the amount of information, a method may be used in which coefficients below a certain threshold are regarded as zero and the number of coefficients to be quantized is reduced.

Problems to be Solved by the Invention However, in conventional methods, when the correlation between pixels is low, it is difficult to limit the quantization range to a sufficiently small size, and quantization cannot be performed efficiently.

In view of this, it is an object of the present invention to provide a highly efficient coding method for video signals by limiting orthogonal transform coefficients for quantization to a smaller range than in the past.

Means for Solving the Problems The present invention is a method of dividing a screen into parallelogram blocks and encoding each block, and before performing orthogonal transformation on the signal within the block, predictive encoding is performed within the block. This is a high-efficiency video signal encoding method characterized by performing orthogonal transformation on the difference signal, and at this time, the prediction direction within each block is determined by changing the prediction direction within each block to one of the blocks forming the boundary. This is a high-efficiency coding method for video signals that is characterized by setting the prediction direction parallel to one side and in the same direction for each column or row.The prediction direction within each block is determined from among several predetermined types. , is a high-efficiency coding method for video signals, which adaptively selects the prediction error in one block by comparing the prediction errors found for each block. This is a highly efficient encoding method for a video signal, characterized in that when the prediction direction is vertical, the horizontal direction is selected, and when the prediction direction is horizontal, the vertical direction is selected.

Effect of the Invention The present invention allows orthogonal transform coefficients to be concentrated more than ever before, and more efficient encoding can be performed using the method described above.

Embodiment FIGS. 1(a) and 1(b) are block diagrams of an encoder and a decoder used in a high-efficiency encoding method for video signals in a first embodiment of the present invention.

Figure (a) is a block diagram of the encoder, where 1 is a buffer memory, 2 is an intra-block DPCM encoder, 3 is an orthogonal transformer, 4 is a quantization range setter, 5 is a quantizer, and 8 is a partial decoder. It is a vessel. The partial decoder 6 includes a quantizer 7, a coefficient matrix restorer 8, an orthogonal transformer 9, a DPCM decoder 10. The DPCM encoder 2 includes a buffer memory 11, and the DPCM encoder 2 includes a delay circuit and a prediction coefficient multiplication circuit.

In this embodiment, the screen is first divided into blocks each having 8 pixels vertically by 8 pixels horizontally. Pixels within one block stored in the buffer memory 1 are first predictively encoded using a DPCM encoder 2 in a predetermined prediction direction. Here, the prediction direction is set parallel to the lateral direction as shown in FIG. 2 (a-1). The first pixel in each column in a block is predictively encoded using adjacent pixels in the previous block that have already been encoded and decoded as predicted values. For the block at the left end of the screen, the pixels of the previous block cannot be used for prediction, so the average value of all pixels in the block is used as the predicted value and the difference is sent. When expressed in terms of two-dimensional frequency characteristics, such a difference value is considered to be concentrated in a high frequency portion in the horizontal direction as shown in FIG. 2 (a-2).

Similarly, the frequency characteristics of the vertical difference values as shown in FIG. 2(b-1) are considered to be concentrated in the high frequency portion in the vertical direction as shown in FIG. 2(b-2).

The difference signal is input to the orthogonal transformer 3 and subjected to two-dimensional orthogonal transformation. The converted coefficients are sent to the quantization range setter 4,
The quantization range is limited to a preset quantization range or a quantization range set for each block by a method such as scanning.

As previously shown in Figure 2 (a-2), the coefficients in this case are considered to be concentrated in the high frequency components in the horizontal direction, so here the range of quantization is limited to the high frequency components as shown in Figure 3. (lower right corner) is one vertex, and is limited to the smallest rectangular part surrounded by rows and columns where all coefficients are zero, and an address indicating the vertical and horizontal range is sent. Coefficients outside this range are zero and are therefore not sent to the quantizer 5.

The coefficients limited by such a procedure are quantized by the quantizer 5. The quantized coefficients are output and sent to the partial decoder 8, where they are decoded in the reverse procedure of encoding. The decoded signal is stored in the buffer memory 11 and used for predicting end pixels when encoding the next block.

It can also be used as a prediction signal for the next frame when performing interframe coding.

FIG. 1(b) shows the configuration of the decoder in this embodiment. In the figure, 7 is a quantizer, 8 is a coefficient restorer, 8 is an orthogonal transformer, 10 is a DPCM decoder, 11
is the buffer memory.

Note that in this example, the pixels of the previous block (the block immediately above in the vertical direction) were used as predicted values for encoding the first column (row in the vertical direction) in the block; The average value of all pixels within the range may be calculated and this may be used as the predicted value. Furthermore, instead of the average value of all pixels, the average value of each row or the average value of each column can be used. This allows encoding to be performed independently within a block.

FIG. 4 is a diagram showing a second embodiment of the present invention.

This is an example in which zigzag scanning is used in limiting the quantization area in the first embodiment.

As explained in the first embodiment, the coefficient in this case is considered to be concentrated in the high frequency component in the horizontal direction (see FIG. 2 (a-2)), so the high frequency component of the coefficient is (lower right corner), perform a zigzag scan to find the last non-zero coefficient, and quantize up to this part.

For quantization, a method is used in which the number of zero coefficients and the magnitude of non-zero coefficients are quantized together.

FIGS. 5(a), 5(b) and 5m are views showing a third embodiment of the present invention.

This is an example in which direction-dependent scanning is used in limiting the quantization area in the first embodiment.

Perform predictive encoding in the horizontal direction (see Figure 2 (a-1))
In this case, as explained in the first embodiment, the orthogonal transform coefficients in this case are concentrated in the high frequency and wave components in the horizontal direction (Fig. 2 (a-
2)), as shown in Figure 5(a), we scan in the vertical direction from the high frequency component of the coefficients (lower right corner), find the last non-zero coefficient, and perform quantum analysis up to this part. become For quantization, a method is used in which the number of zero coefficients and the value of non-zero coefficients are combined and unquantized.

With this method, the number of required bits can be kept small when encoding in combination with the number of zero coefficients.

Also, prediction is performed in the vertical direction (see Figure 2 (b-1))
In this case, the coefficients are concentrated in the high frequency components in the vertical direction (Fig. 2 (
b-2)) Therefore, as shown in Figure 6(b), we scan in the horizontal direction from the high frequency component of the coefficients (lower right corner), find the last non-zero coefficient, and scan up to this part. Quantize.

FIGS. 6(a) and 6(b) are block diagrams of an encoder and a decoder used in a high-efficiency encoding method for video signals according to a fourth embodiment of the present invention. In this embodiment, the prediction direction is switched based on the accumulation of prediction errors. The encoder shown in FIG. 1(a) includes a prediction direction determiner 1 including a prediction error accumulator and a comparator in addition to the encoder in the first embodiment (see FIG. 1(a)).
2 has been added. These prediction error accumulation circuit, comparator θ, etc. can all be realized with a simple configuration.

In this embodiment, the prediction error accumulator is used to calculate 1
Prediction errors within a block are accumulated, the comparator judges the magnitude of the errors, and the smallest prediction direction is selected. At the same time as outputting the determination code obtained from this, a decoding output corresponding to the selected prediction direction is sent. Note that the accumulation of prediction errors used to determine the prediction direction includes, for example, 1 of the absolute value of the prediction error.
All you have to do is take the total sum within the block and determine its size.

The buffer memory 11 includes a nonlinear quantization circuit 2 for each prediction direction.
One block's worth of code words obtained from the above are stored, and when a judgment code is received, one block's worth of code words obtained from the prediction direction indicated by this code is output. The prediction direction determination code (in this embodiment, 1 bit in two directions) is placed at the beginning of each block.

Note that in this embodiment, two types of prediction directions are set, but the number and direction of these prediction directions are arbitrary.

In addition, although we used the accumulation of prediction errors to determine the prediction direction,
This may be an accumulation of difference values. Further, although the accumulation of prediction errors is determined by the sum of absolute values, the sum of the squares of prediction errors or the square root thereof may also be used.

FIG. 5(b) is a block diagram of the decoder in this embodiment. In consideration of decoding, after the decoding circuit (quantizer 7, coefficient restorer 8, orthogonal transformer 9, DPCM decoder 10) performs decoding calculations for one block, the prediction direction setter 13 The order is rearranged according to the judgment code from among the directions placed.

The present invention can be configured in various ways other than those shown in the above embodiments.

For example, in the above embodiment, the blocks are divided into 8×8 pixels, but the size of the blocks is arbitrary. Further, although the quantization method uses a method of quantizing the number of zero coefficients and the value of non-zero coefficients together, this method does not depend on the quantization method.

As described in detail, according to the present invention, orthogonal transform coefficients can be concentrated more than before, and more efficient encoding can be performed.

[Brief explanation of drawings]

FIGS. 1(a) and 1(b) show an encoder (a) for explaining a high-efficiency encoding method for video signals in the first embodiment of the present invention.
), Block diagram of decoder (b), Figure 2 (a-1) (a-
2) (b-1) (b-2) is the prediction direction (a-1) (b-1) in the block in the same example and the distribution of frequency components of the difference signal (a-2) (b-2 ), FIG. 3 is a diagram showing a method for determining the quantization range of orthogonal transform coefficients in the same embodiment, and FIG. A diagram showing a method for determining the quantization range of transform coefficients, Figure 5 (a) (
b) is a diagram showing the method for determining the quantization range of orthogonal transform coefficients in the third embodiment of the present invention ((a) is in the vertical direction, (b) is in the horizontal direction); ) is a block diagram of an encoder (a) and a decoder (b) for explaining the high-efficiency encoding method of a video signal in the third embodiment of the present invention, and FIG. 7 is an explanatory diagram of an orthogonal transform coefficient matrix. , and FIGS. 8 to 10 are diagrams showing a conventional method for determining the quantization range of orthogonal transform coefficients. 200. Intra-block DPCM encoder, 311. Orthogonal transformer, 12. ,,Prediction direction determiner. Name of agent: Patent attorney Shigetaka Awano Haka 1st person! Figure Figure Figure Figure Figure (CL) Tumor Map (α-T) Scararo νri (α-2) Figure Figure Country 1km Wave Kashuo Figure Figure! O diagram

Claims (4)

[Claims] (1) In a method of dividing the screen into parallelogram blocks and encoding each block, before performing orthogonal transformation on the signal within the block, predictive encoding is performed within the block, and the difference signal is A highly efficient encoding method for a video signal, characterized by performing orthogonal transformation on the video signal. (2) Claim 1 characterized in that the prediction direction within each block is set parallel to one vertical or horizontal side forming the boundary of the block, and in the same direction for each column or row.
A highly efficient encoding method of a video signal as described above. (3) The prediction direction within each block is adaptively selected from several predetermined types by comparing the cumulative prediction error within one block obtained for each type. A highly efficient encoding method for a video signal according to claim 1. (4) The scan direction of orthogonal transform coefficients in each block is
3. The high-efficiency encoding method of a video signal according to claim 2, wherein when the prediction direction is vertical, the horizontal direction is selected, and when the prediction direction is horizontal, the vertical direction is selected.