JPS621383A - System and device for detecting scene change of animation picture signal - Google Patents

Abstract

PURPOSE:To detect accurately a scene change by adding the evaluated results of the motion compensation forecasting error in plural lines and comparing the addition result with a determined threshold. CONSTITUTION:A variable delay circuit 13 delays variably the picture signal of one picture before outputted from a frame memory 12 in accordance with a dynamic vector supplied through a line 113 and outputs the signal to a subtractor 14 as a motion compensation forecasting value. The subtractor 14 subtracts the motion compensation forecasting value from the delayed input signal supplied through a line 1814 to obtain the motion compensation forecasting error. An absolute value calculator 15 calculates the absolute value of the motion compensation forecasting error and outputs it to an integrating circuit 16. The integrating circuit 16 adds the motion compensation forecasting error from the start of a frame and outputs the result to a comparator 17. When the addition result exceeds the preliminarily determined threshold, the comparator 17 discriminates the scene change to output 1 to an output terminal 2000.

Description

DETAILED DESCRIPTION OF THE INVENTION (Field of Industrial Application) The present invention relates to digital transmission technology for television signals, and particularly to technology for detecting significant changes in content between screens.

(Prior art) Conventionally, as a means of detecting a scene change, one frame of a past screen is stored in a frame memory, a frame difference value is calculated using this, and this absolute value is added by an appropriate line. There is a known technology that determines that a scene change has occurred when the value exceeds a threshold (Miyahara et al., ``Measurement and Study of Frame Difference Signal Characteristics for Television Signal Band Compression'' JNHK Technical Research, Volume 29, No. 1, PP, 33-PP, 49, 1972).

(Problem to be Solved by the Invention) However, the above conventional technology uses frame difference values, so even if the scene does not change, the absolute value of the frame difference value increases for images with large movements such as panning. Therefore, there was a drawback that this could be erroneously detected as a scene change.

(Means for solving the problem) According to the present invention, in detecting a scene change,
By adding the evaluation results of motion compensation prediction errors across multiple lines and comparing the addition results with a predetermined threshold,
A scene change detection method characterized by detecting a scene change is obtained.

Further, according to the present invention, there is provided a first storage means for storing at least one frame of an input moving image signal, and a block consisting of a plurality of pixels using the moving image signal output from the first storage means and the input moving image signal. means for calculating a motion vector in units or pixel units; second storage means for storing the input video signal and capable of changing the amount of delay according to the delay time indicated by the motion vector; and an output of the means, that is, motion compensation. means for calculating the difference between the predicted value and the input video signal, that is, a motion compensated prediction error value; means for adding evaluation results for the motion compensated predicted value for a plurality of lines; A scene change detection device for a moving image signal is realized, which is characterized by having a means for comparing sizes.

(Operation) First, in the present invention, it is necessary to obtain a motion compensation prediction error value.

The motion compensated prediction error can be calculated using, for example, a motion vector detected by a block matching method. As shown in Figure 2, the block matching method involves storing a past screen y in a frame memory, and then selecting a block A' that has the highest correlation with block A of the current frame F from the stored past screen. , and the displacement of the positions of both within the screen is taken as the motion vector V. The motion compensated prediction error corresponds to a difference value for each pixel between two blocks associated by a motion vector. As an evaluation criterion for the height of correlation, minimizing the square of the difference value between corresponding pixels or the total sum of absolute values within a block is used.

There is no direct relationship between the present invention and the method of determining motion compensation prediction error.
The motion compensated prediction error may be obtained by a method other than the above.

FIG. 1 is a diagram showing how the motion compensation prediction error and the variance value of the inter-frame difference value change with respect to the speed of inter-frame movement and panning speed.

The variance value of the frame difference value is expressed by the formula (2), and increases monotonically as the movement becomes faster. On the other hand, the variance value of the motion-compensated prediction error is expressed by the equation 0, and when the motion is an integer multiple of the pixel interval, the motion-compensated prediction is accurate, so the prediction error is zero as shown in the figure. Become.

Here, σ2 and 2.0MC2 represent the variance values of the frame difference value and the motion compensated prediction error value, respectively, σo2 is the average power of the image, and α is a parameter corresponding to the complexity of the screen.

σ2=2a2(1-e-01)...
・・■n'ts a" = 2o2 (1-v-"-")"'
・・・・・・■MCo→ [V] represents the integer part of V. However, in the case of an image where there is no correlation between frames such as a scene change, both the inter-frame difference value and the variance value of the motion compensation prediction error are the second This value is close to the upper limit of the variance value for the difference value between picture frames. Therefore, when using interframe difference values, it is not possible to distinguish between scene changes and fast-moving images, but by using motion compensation prediction error and comparing the magnitude with the judgment threshold as shown in Figure 2, it is possible to distinguish between scene changes and fast-moving images. Only scene changes can be detected accurately.

(Example) Next, the present invention will be explained in detail using FIG.

A moving image signal input from the input terminal 1000 is supplied to the D converter 10. The human converter 10 converts the digitized moving image signal into lines 1011 and 1012°10.
18 to the motion vector detection circuit 11, frame memory 12, and delay circuit 18, and outputs an F pulse representing the timing of the beginning of each screen to the addition circuit 16 via line 1016.

The motion vector detection circuit 11 calculates a motion vector for each block or pixel by using the video signal supplied from the line 1011 and the past video signal supplied from the line 1211. The calculated motion vector is output to line 1113 as the optimal prediction method. This motion vector detection method is described, for example, by Ninomiya in [Motion Correction in Interframe Coding], Journal of the Institute of Electronics and Communication Engineers, VOL.
J64-B No., 1975, pp. 24-31) can be used.

The frame memory 12 is capable of storing moving image signals for about one screen, and outputs image signals from about one screen time ago to lines 1211 and 1213.

The variable delay circuit 13 variably delays the moving image signal supplied via the line 1213 according to the optimal prediction method supplied via the line 1113, and outputs the video signal supplied via the line 1213 as a motion compensated predicted value.
14 to the subtracter 14. The delay circuit 18 is
The motion vector detection circuit 11 calculates a motion vector of the input signal, the variable delay circuit 13 delays the input signal by a time necessary to generate a motion compensation predicted value, and outputs the signal to the subtracter xS14. Subtractor 14 subtracts the motion compensated prediction value from the delayed input signal provided via line 1814 to obtain a motion compensated prediction error and outputs it to absolute value calculator 15 . Absolute value calculator 15
takes the absolute value of the motion compensated prediction error and outputs it to the integration circuit 16 via line 1516. Integration circuit 16 connects line 10
Reset with the frame head information supplied from 16,
It begins summing the motion compensated prediction errors provided via line 1516 and outputs the result to comparator 17. Comparator 17
When the motion compensation prediction error added from the beginning of the frame exceeds a predetermined threshold, it outputs 1 to the output terminal 2000, indicating that a scene 6 change has occurred.

(Effects of the Invention) According to the present invention, 1. A scene change detection method with fewer detection errors has been realized for various images, and a method for instantaneously suppressing information is provided at the scene change, when the largest amount of information is generated in compression coding of video signals. It is of great significance to put the present invention into practical use.

[Brief explanation of the drawing]

Figure 1 is a diagram showing the relationship between the frame difference value, the variance value of motion compensation prediction error, and the speed of motion, Figure 2 is a diagram showing the principle of motion vector detection using the block matching method, and Figure 3 is FIG. 2 is a diagram representing an embodiment of the present invention. In the figure, 1000 is the input terminal, 10 is the 0 converter, and 11 is the input terminal.
1 is a motion vector detection circuit, 12 is a frame memory, 13 is a variable delay circuit, 14 is a subtracter, 15 is an absolute value calculator, 1
6 is an integration circuit, 17 is a comparator, 18 is a delay circuit, 200
0 represents the output end.

Claims (2)

[Claims] (1) When detecting a scene change in a moving image signal such as a television signal, the motion compensation prediction error evaluation results are added over multiple lines, and the scene change is detected by comparing the addition result with a predetermined threshold. This is a scene change detection method that is characterized by the following. (2) a first storage means for storing at least one frame of an input moving image signal; using the moving image signal outputted from the first storage means and the input moving image signal in units of blocks consisting of a plurality of pixels or in units of pixels; means for calculating a motion vector; second storage means for storing the input video signal and capable of changing the amount of delay according to the delay time indicated by the motion vector; an output of the means, that is, a motion compensated predicted value; means for calculating a difference from the input video signal, that is, a motion compensated prediction error value; means for adding evaluation results for the motion compensated predicted value for a plurality of lines; and means for comparing the added value with a predetermined threshold value. What is claimed is: 1. A scene change detection device for a moving image signal, comprising: