JPS6058636B2 - Interframe coding method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an interframe encoding method for band compression encoding a television signal.

Conventional interframe coding methods such as the conditional pixel rewriting method and the composite difference method use interframe difference coding with the pixel at the corresponding position one frame before, or pre-predictive coding of the interframe difference. We have been performing band compression encoding of television signals. With these methods, when there is little movement on the screen, the amount of information generated is small and the television signal can be transmitted with good image quality, but when the movement is intense, the prediction error becomes large and the compression efficiency is poor, resulting in poor quality. Image quality cannot be maintained. In order to solve this drawback, the conventional technology uses the following method.That is, by dividing the frame of the television screen into blocks each containing an appropriate number of nxn pixels, the movement of the image is If viewed locally within a block, it can be regarded as parallel movement. Therefore, considering the movement of a pixel translation to the right per frame as shown in Figure 1, the predicted value Zij of pixel Zij can be expressed as Zij yi - a , j, it is possible to improve the preliminary efficiency compared to the normal interframe coding using Zij ■ yij in the conventional method.
This method has the disadvantage that prediction efficiency is low when the movement is a non-integer number of pixels per frame. In order to solve these drawbacks of the prior art, the present invention improves the band compression rate and image quality by encoding the television signal using an adaptive prediction method that follows the movement of the television screen. The present invention provides an interframe coding method that can achieve the following.

The present invention will be explained in detail below.

The present invention provides a block consisting of a plurality of pixels of the current frame, a corresponding block consisting of the same number of pixels as the plurality of pixels of an adjacent or adjacent preceding frame located at the same position, and a block that is obtained by vertically and horizontally moving this block in parallel. The squared error (
(or absolute value error) is calculated, and blocks with smaller values are defined as similar blocks.

Several similar blocks are selected in descending order of error, and prediction is performed using a weight vector determined from the moving direction and amount of movement from the corresponding block, and the inter-frame error intra-block average value. That is, as shown in FIG. 2, blocks 1, 2, . . . m similar to the block in the current frame including the pixel Zij are selected in the order of similarity in the previous frame.

The squared error of each block is ε0)×E(″ゝ.Weight vector W=(WO), w(2), . . ., w determined corresponding to the similarity of each block. (′″)
), and the pixel Yij (. , is determined by solving a linear equation consisting of a correlation matrix so as to minimize the intra-block squared error between frames.

Here, '3'w(k)=1.

k=1 By using the prediction function expressed by the linear combination shown in Equation (1), it is possible to increase the prediction accuracy for images with small movements, enhance the compression effect, and improve the image quality.

Furthermore, by taking four similar blocks centered on the most similar block with the smallest inter-frame error and applying the above method, it is possible to find the amount of movement in which the movement is a non-integer pixel per frame using the weight vector. For example, when predicting the pixel Zij of the current frame based on the arrangement of the pixel Yij of the previous frame as shown in FIG.
It is expressed as a vector adding the distance to the intersection P connecting the points internally divided by the weighting coefficient w(k) (k=1...4).
Furthermore, if similar blocks and weight vectors are found for each block of the current frame and prediction is performed adaptively, the method can be applied to images with complex movements other than parallel movement. Fourth
The figure shows an embodiment of the present invention, in which 1 is a signal input terminal, 2 is a blocking circuit, 3 is a memory, 4 is a similar block detection circuit, 5 is a frame memory, 6 is a prediction coefficient determination circuit, and 7 is a prediction circuit. , 8 is a subtracter, 9 is a quantizer, 10 is an adder,
11 is a buffer memory 18 12 is a signal output terminal.

First, a digitized television signal is input through a signal input terminal 1, divided into blocks each consisting of N×n pixels in a blocking circuit 2, and simultaneously stored in a memory 3 until similar blocks and prediction coefficients are determined.

Similar block detection circuit 4 calculates the intra-block difference average using the output of blocking circuit 2 and the pixel values of the previous frame stored in frame memory 5 to find m similar blocks. Next, the prediction coefficient determination circuit 6 uses the frame difference average value to determine the prediction coefficient W(
6) is calculated, and the prediction circuit 7 calculates the predicted value 21j of each pixel value in the block using this coefficient. The predicted value is subtracted from the corresponding pixel Zij stored in the memory 3, and the result is quantized in the quantizer 9. The quantized output is locally decoded by adding it to the predicted value from the prediction circuit 7 in the adder 10, and then stored in the frame memory 5.
to be memorized. At the same time, the buffer memo is led to the signal output terminal 12 via the river 1. As explained above, the present invention uses an adaptive prediction method that follows motion to create a coding method with high prediction efficiency and excellent image quality. It can be applied to the conversion method.

[Brief explanation of drawings]

Figure 1 is a diagram for explaining the conventional method effective for interframe prediction when the image movement is one integer pixel frame, and Figure 2 is for explaining the principle of linear combination prediction for moving images according to the present invention. FIG. 3 is a diagram for explaining the detection of the amount of movement in which the amount of movement per frame is a non-integer pixel as one of the application examples of the present invention, and FIG. 4 is a diagram according to the present invention. FIG. 2 is a block diagram showing a configuration sequence of an interframe band compression encoding device. 1...Signal input terminal, 2...Blocking circuit, 3.
...Memory, 4...Similar block detection circuit, 5...
Frame memory, 6... Prediction coefficient determination circuit, 7...
prediction circuit, 8... subtracter, 9... quantizer, 10.
...Adder, 11...Buffer memo ■812...Signal output terminal.

Claims (1)

[Claims] 1. A block of an appropriate shape containing a plurality of the same number of pixels is set at spatially corresponding positions in two sequentially adjacent or close frames of a television signal, and one of the two frames is set. An interframe error is calculated between a block in the current frame, a corresponding block in the other preceding frame, and a plurality of blocks obtained by vertically and horizontally moving the corresponding block in the preceding frame, and Select m blocks as similar blocks in the preceding frame in descending order of the intra-block integral value of the inter-frame error, and assign a prediction coefficient determined by the integral value of the inter-frame error to each similar block. For each pixel included in the block of the current frame, the prediction coefficient given to m blocks from the pixel values of m pixels at corresponding positions in the m similar blocks is defined as a coefficient. An interframe encoding method that performs prediction using a linear combination of