JPH07184206A - Picture coder - Google Patents

Abstract

PURPOSE:To obtain the scene adaptive prediction coding system characterized in that the transmission efficiency is high and deterioration in a picture is less even at frequent scene changes for a moving picture signal. CONSTITUTION:An input picture signal is given to a subtractor section 1, in which the signal is subtracted from a predicted value from an adaptive prediction section 20 and the resulting signal is transmitted. An in-frame predict section 6, an inter-frame predict section 7 and a frame memory 22 writing a signal by a scene change detection section 21 are connected in parallel to the adaptive predict section 20. Power of the input signal and that of respective predict error data are compared and a predicted value with least power is outputted as a predicted value suitable for the input signal. Picture data corresponding to a scene change are stored by plural frames and used for prediction or a pattern predicted in advance is provided and used for the prediction, then efficient transmission with less difference component to be transmitted in the case of even a large scene change as well as a picture comprising similarly repeated scenes and deterioration in picture quality is reduced in the case of transmission at processing speed.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an interframe predictive coding apparatus for efficiently coding a television signal used in a moving picture transmission apparatus, a recording apparatus or the like.

[0002]

2. Description of the Related Art An interframe predictive coding system is often used as a high-efficiency coding system used for recording and transmitting moving images. This predictive coding method uses a feature that the inter-frame correlation of the television signal is high and the inter-frame difference signal is small, and the difference is calculated by using the image data of one frame before the input image data as the prediction signal. In this method, the inter-frame difference signal is compressed as it is encoded as prediction error data. Conventionally,
An encoding system of this kind is shown in FIG. This figure is a block diagram of an encoder of a conventional compressed image data recording / reproducing system disclosed in Japanese Patent Laid-Open Publication No. 64-10793, in which 1 is a subtractor, 2 is a quantizer, and 3 is Code conversion circuit, 4 dequantizer, 5 adder, 6 intra-frame prediction circuit, 7 inter-frame prediction circuit, 8 background prediction circuit, 9 prediction mode control circuit, 10
Is a local decoder, and 11 is a predictor.

Next, the operation will be described. The input image data is subtracted from the prediction signal in the subtractor 1 and supplied to the quantizer 2. One of the prediction error data converted from the level value to the level number in the quantizer 2 is converted into an unequal length code in the code conversion circuit 3. The other is fed back to the subtractor 1 via the local decoder 10. The local decoder 10 supplies the output of the inverse quantization unit 4 which inversely processes the quantization, that is, the level number is inversely converted into the level value, to the subtractor 1 through the predictor 11, while the input side of the predictor 11 is supplied. Positive feedback is provided to the adder 5 provided in the. In the present example, the predictor 11 uses the intra-frame prediction circuit 6 and the inter-frame prediction circuit 7.
Also, the background prediction circuit 8 is connected in parallel, and the output of one of the prediction circuits is selectively connected to the subtractor 1 by a changeover switch. This is because there is no correlation between the image signal of the previous frame and the image signal of the current frame when there is a scene change (scene change) or violent movement, so it is better than inter-frame predictive coding. This is because it is more efficient to perform intra-frame predictive coding using intra-correlation and background predictive predictive coding using correlation with a background image. In the case of the present example, the prediction mode control circuit 9 operates according to the scene change information or the like by the power comparison of the prediction error data or the operator's judgment,
I am trying to switch. The background prediction circuit 9
For example, it is sufficient to remove all moving subjects and store the captured image, but in general, the background image should be gradually updated in consideration of the case where the background data changes in the middle due to changes in lighting conditions. The following methods have been proposed for generating all of them. Figure 9: "Multidimensional signal processing of TV images" by Takahiko Fukibuki
It is a background prediction circuit reported in p208-209.
Using the difference signal Y-Y _F between frames where a large difference divides the place moving area and the difference is small to the stationary area,
In the stationary area, this is regarded as a background image and gradually corrected and updated. Looking at the polarity ε of the difference between the input image signal Y and the background signal, a slight value 1 / k is added to the background signal to bring them closer to each other. Therefore, when the prediction mode control circuit 9 determines that the prediction by the background prediction circuit 8 can suppress the power of the prediction error data most, the background prediction mode is selected.

[0004]

Since the conventional predictive coding system is constructed as described above, even if the background predicting circuit is provided, the same background does not always appear. In addition, since the background prediction cannot support scene changes in which the image signal changes rapidly and a large amount of prediction error data is generated, it is difficult for the data amount to be constant, and quantization is required when transmitting at a constant data rate. However, there is a problem in that the deterioration of the transmitted image becomes large because it is necessary to roughen the frame and to drop the frame to be transmitted.

The present invention has been made in order to solve the above-mentioned problems, and the prediction error of inter-frame prediction or intra-frame prediction error in a scene change of an image signal. The purpose is to suppress the sudden occurrence and reduce the deterioration of the image when transmitting at a constant speed.

[0006]

A predictive coding apparatus according to a first aspect of the present invention has at least one second frame memory controlled by a scene change detection signal, and the frame The output of the memory is used as a prediction signal.

Further, in the predictive coding apparatus according to the second aspect of the present invention, the predictive coding apparatus has a frame average value calculation unit for a frame in which a scene change is detected, and controls the writing of the frame memory according to the magnitude of the average value. It is something to do.

In the predictive coding apparatus according to the third aspect of the present invention, the output of the second frame memory is low-pass filtered and the output is used as a predictive signal.

Further, in the predictive coding device according to the fourth aspect of the present invention, there is provided an average value calculation unit for the frames in which the scene change has occurred, and this average value is used as the prediction signal.

Further, in the predictive coding device according to the fifth aspect of the present invention, there is provided at least one third frame memory in which the predicted signal is written as a pattern, and this output is used as the predicted signal. To do.

Further, in the predictive coding apparatus according to the sixth aspect, the unit of detecting a scene change and writing the second frame memory is performed not in a frame unit but in an arbitrary unit, for example, a block unit, and outputs the memory. Alternatively, adaptive prediction is performed in block units using the above average value as a prediction signal.

Further, in the predictive coding device according to the seventh aspect, the scene change of the block is detected by using the result of the scene change judgment of the adjacent surroundings.

[0013]

The predictive coding apparatus according to the first aspect of the present invention has the second frame memory written by the scene change detection signal, and the first image signal that causes the scene change is written in the frame memory. . This image is similar in a scene change that occurs later, for example, when switching multiple surveillance cameras and transmitting images, the same camera image signal in the past becomes the image with the highest correlation, so the difference can be reduced. As a possible prediction signal, the first image signal of the past scene change is output from the second frame memory to the adaptive prediction unit.

The predictive coding apparatus according to the second aspect of the present invention limits the writing of the second frame memory according to the magnitude of the average value of the frames in which the scene change is detected. Since many are not written, it is possible to support more scenes with a smaller number of memories.

The predictive coding apparatus according to a third aspect of the present invention performs low-pass filter processing on the output of the second frame memory to obtain an image in which an input image signal is written. Therefore, even if the camera is blurred or shakes, it can be dealt with.

In the predictive coding apparatus according to the fourth aspect of the present invention, in many cases, the first one frame of the moving image starts from the screen of all white and all black. Therefore, the average value of the scene-changed screen is obtained, and this average value is output to the adaptive prediction unit as a prediction signal.

Further, in the predictive coding apparatus according to the fifth aspect, a pattern in which the upper half is bright and the lower half is dark, a pattern in which the left half is gradually brighter, and the like are changed depending on the image signal which is likely to be input. A certain number of pattern signals are written in advance and output as a scene change prediction signal to the adaptive prediction unit.

Further, in the predictive coding apparatus according to the sixth aspect, the unit of scene change detection and the writing of the second frame memory is divided not into frame units but into smaller block units, etc. This is a prediction signal more suitable for a scene change in response to the change.

Further, in the predictive coding apparatus according to the seventh aspect, when the conditions for detecting a scene change are delicate, the fact that the background is present in a certain amount is used and the surrounding judgment result is also used to determine the scene. A change is detected.

[0020]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT An embodiment of the present invention will be described below with reference to FIG. Example 1. FIG. 1 is a block diagram of an encoding apparatus showing an embodiment of claim 1 of the present invention.
4, 5, 6, and 7 are the same as the conventional device shown in FIG. 20 is an adaptive prediction unit, 21 is a scene change detection unit, 22 is a plurality of scene change frame memories, 2
3 is a frame memory control unit, and 24 is a local decoding unit.

Next, the operation of the embodiment of the present invention will be described. The input image data is subtracted from the prediction signal in the subtractor 1 and supplied to the quantizer 2. Subsequent processing is the same as the processing of the conventional example, and is omitted. The other input image data is taken as a difference from the output of the frame memory of the scene change detection section 21 in the scene change detection section 21. The absolute value of the difference value is added for one frame, and if the value is larger than an arbitrary value, it is determined as a scene change. If it is judged as a scene change,
The first 1 immediately after the scene change in the frame memory controller
The frame is written in the scene change memory 22. The frame memory control unit sequentially writes to one of the plurality of scene change memories 22 each time a scene change is detected. In the case of this example, the outputs of the intra-frame prediction circuit 6, the inter-frame prediction circuit 7, and the plurality of scene change memories 22 are connected in parallel to the adaptive prediction unit 20,
By comparing the power of the input image data and the respective difference data, the output with the lowest power is selected as the prediction signal and output to the subtractor 1. A number indicating which scene change memory the output is from and a quantized prediction error are transmitted to the receiving side. In the case of normal input image data, either the intra-frame prediction circuit 6 or the inter-frame prediction circuit 7 is selected, but when a similar background image is repeatedly input, it is written in the scene change memory. By selecting the background image, it is possible to reduce the prediction error and transmit the image. Furthermore, when there is a scene change, intraframe predictive coding is performed rather than normal interframe predictive coding because the image data of the previous frame and the image data of the current frame have no correlation. If there is a more optimal change frame memory output, that value can be used as a prediction signal, so that it is possible to reduce the prediction error and transmit. In particular, when switching between multiple surveillance cameras to transmit a moving image, the images input from each camera are almost the same, so it is possible to detect a scene change when switching between cameras and store the images. By selecting the same past camera output accumulated at the time of the next camera switching as the prediction signal, it is possible to reduce the occurrence of a prediction error at the scene change.

By writing a predictable signal in the scene change frame memory as an initial value in advance, it is naturally possible to perform efficient transmission with a small difference value from the start of image transmission. Is.

In the above embodiment, the first start image after the scene change is written in the scene change memory 22, but the same effect can be obtained by writing the last image immediately before the scene change. Obtainable.

Example 2. 2 is a block diagram of an image coding apparatus showing an embodiment of the present invention, and an average value calculation unit 25 is added to the block diagram of the image coding apparatus showing an embodiment of the first claim of FIG. It was done. In the first embodiment described above, the plurality of scene change memories 22 are rewritten in order every time a scene change is detected. However, the average value of each frame is calculated by the average value calculation unit 25, and the average value is calculated in advance according to the average value. It is written into the determined scene change memory 22. By determining the frame memory to be written based on the average value, only the image frames of similar scenes are not written, so that the number of scene change memories 22 can cope with more scene changes.

Example 3. FIG. 3 is a block diagram of an image coding apparatus showing an embodiment of the present invention, and a low-pass filter 26 is added to the block diagram of the image coding apparatus showing an embodiment of the first claim of FIG. It was done. In the above-described first embodiment, the outputs of the plurality of scene change memories 22 are used as they are each time a scene change is detected. However, by outputting the output passed through the low pass filter to the adaptive prediction unit 20, It can also be used when a scene-changed image causes camera shake or blurring.

Example 4. FIG. 4 is a block diagram of an image coding apparatus showing an embodiment of the present invention.
4, 5, 6, and 7 are the same as the conventional device shown in FIG. Reference numeral 20 is an adaptive prediction unit, and 24 is a plurality of pattern frame memories.

Next, the operation will be described. For example, when a landscape is photographed by a camera, the upper half corresponding to the sky may be white and the lower half corresponding to the ground may be dark if only the luminance signal is considered. When a mountain is photographed against the sky, a mountain-shaped dark pattern can be formed on a light background. For this reason,
As described above, some pattern signals that change in brightness and darkness and are likely to be input in advance are written in the memory. When a pattern signal having a high correlation with the first frame in which a scene change has occurred is written in the memory 24, the pattern signal is selected by the adaptive prediction unit 20, and therefore a prediction error with the selected pattern signal is transmitted. As a result, the prediction error signal to be transmitted can be reduced. Further, when the input image is already known as a background or the like for monitoring or the like, the image is written in the pattern frame memory 24 in advance, and the output image of the frame memory 24 is used as a prediction signal for transmission. It is possible to reduce the prediction error signal. The adaptive prediction unit 20 compares the powers of the respective difference data between the input pattern output of the plurality of pattern frame memories 24 and the input image data, and selects the pattern output with the least power as the prediction signal and the subtractor. It is output to 1.
Prediction error between the input image data and the pattern output is calculated in the subtracter 1, and the prediction error data converted from the level value to the level number in the quantizer 2 is transmitted to the receiving side together with the number indicating the pattern output obtained by the difference. To be done.

Example 5. FIG. 6 is a block diagram of an image coding apparatus showing one embodiment of the present invention.
2, 4, 5, 6, and 7 are the same as the conventional device shown in FIG. Reference numeral 20 is an adaptive prediction unit, 21 is a scene change detection unit, and 25 is an average value calculation unit.

Next, the operation will be described. When the scene change detection unit 21 detects a scene change, the average value calculation unit 25 calculates the average value of the frames in which the scene change is detected. This average value is used as the adaptive prediction unit 2
By outputting to 0, the adaptive prediction unit 20 outputs all white,
When starting from an all-black screen, the average value is selected as the optimum predicted value. The adaptive prediction unit 20 compares the power of the respective difference data with the input image data, and the output with the lowest power is selected as the prediction signal and output to the subtractor 1. The subtractor 1 calculates the prediction error between the input image data and the prediction signal, and the quantizer 2 outputs the prediction error data converted from the level value to the level number, and the prediction error data is transmitted to the receiving side together with the average value of the differences. It Efficient transmission can be achieved by sending only one average value for a frame.

Example 6. Further, in the above-described embodiment, the sum of absolute difference values of the inter-frame difference signals for one frame is compared with the set value as the scene change detection unit, and the writing process is performed in the second frame memory in frame units. By detecting the scene change in the area-divided block and performing the writing process to the second memory, the memory capacity of the second memory can be reduced and the image change in the first frame of the scene change can be handled in block units. Therefore, more efficient transmission can be performed.

Example 7. Further, as shown in FIG. 7, the setting value is switched according to the number of scene change determination results of the adjacent peripheral blocks output from the memory 36, and the result determined in the vicinity of the set value is further detected by detecting the scene change. Can be done accurately. In the set value table 35 here, the set value is reduced according to the number of scene change detections of adjacent blocks to facilitate determination as a scene change.

In the above description, inter-frame differential encoding is taken as an example of the predictive encoding method using the correlation between image units, but other methods such as motion-compensated inter-frame prediction and inter-field prediction can be used. It is also possible to use
It is also possible to perform transform processing such as DCT (discrete cosine transform) or vector quantization on the prediction error data generated in these coding processes.

[0033]

Since the present invention is constructed as described above, it has the following effects.

When a plurality of frames of past image data corresponding to a scene change are accumulated and used for prediction, when switching several surveillance cameras and transmitting an image signal in which similar scenes are repeated, Of course, even when the scene changes greatly, efficient transmission with a small difference value to be transmitted can be performed, and there is an effect of reducing image deterioration at constant speed transmission.

Further, since the scene change frame memory to be written is determined based on the average value of the frames at each scene change, only image frames of similar scenes are not written, which is small. The frame memory can handle more scene changes.

Further, by performing the low-pass filter processing on the output of the scene change frame memory, it is possible to deal with the case where camera blur or shake occurs in the image in which the input image signal is written.

Further, by using the average value of the scene-changed image as the prediction signal, all white,
Efficient transmission can be performed even for images starting from all black.

Further, by adopting a configuration in which a signal that can be predicted in advance is stored in the memory as a prediction signal in advance, efficient transmission with a small difference value can be performed when the correlation with the image prepared in advance is high.

Furthermore, by adopting a configuration in which scene change detection and adaptive prediction are performed in a frame or block that is divided into small regions, the memory amount can be reduced and more efficient transmission corresponding to image changes within the frame can be performed. it can.

Further, by detecting the scene change using the scene change judgment result of the adjacent peripheral blocks, the result judged near the set value can be more accurately performed.

[Brief description of drawings]

FIG. 1 is a block diagram of a scene adaptive prediction system showing an embodiment of the present invention.

FIG. 2 is a block diagram of a scene adaptive prediction method showing a second embodiment of the present invention.

FIG. 3 is a block diagram of a scene adaptive prediction system showing a third embodiment of the present invention.

FIG. 4 is a block diagram of a scene adaptive prediction system showing a fourth embodiment of the present invention.

FIG. 5 is a diagram of pattern signals used in a fourth embodiment of the present invention.

FIG. 6 is a block diagram of a scene adaptive prediction system showing a fifth embodiment of the present invention.

FIG. 7 is a block diagram of a scene change detecting means showing an embodiment of the present invention.

FIG. 8 is a block diagram showing an encoding unit of a conventional compressed image data recording / reproducing system.

FIG. 9 is a block diagram showing a conventional background prediction circuit.

[Explanation of symbols]

 1 Subtractor 2 Quantizer 3 Unequal Length Coding 4 Inverse Quantizer 5 Adder 6 Intraframe Predictor 7 Interframe Predictor 8 Background Predictor 9 Prediction Mode Control Circuit 10 Local Decoder 11 Predictor 20 Adaptive prediction unit 21 Scene change detection unit 22 Scene change frame memory 23 Frame memory control unit 24 Pattern frame memory 25 Average value calculation unit 26 Low-pass filter 31 Frame memory 32 Subtractor 33 Absolute value cumulative addition unit 34 Comparison unit 35 Setting value table 36 memory It should be noted that the same reference numerals in the respective drawings indicate the same or corresponding portions.

Claims (7)

[Claims] 1. A subtracter for taking a difference between an input image signal and a prediction signal, a quantizer for converting the difference signal output into a level signal, and an inverse unit for inversely converting the level signal into a quantized difference signal output. A quantizer, an adder for adding the quantized difference signal output and the prediction signal, a first frame memory for storing the addition result, and a scene change detection for detecting a scene change of the input image signal. Section, a second frame memory for storing at least one addition result controlled by a scene change detection signal, and an optimum prediction from image data stored in the first and second frame memories. An image coding apparatus comprising an adaptive prediction unit that outputs a signal. 2. The image coding apparatus according to claim 1, further comprising: a frame average value calculation unit for the input image signal, wherein writing into the second frame memory is controlled by the average value. . 3. The image coding according to claim 1, further comprising a low-pass filter at an output of at least one of the second frame memories controlled by the scene change detection signal. apparatus. 4. The image coding apparatus according to claim 1, further comprising an average value calculation unit for frames of the input image signal in which a scene change is detected as the prediction signal. 5. A subtractor for taking a difference between an input image signal and a prediction signal, a quantizer for converting the difference signal output into a level signal, and an inverse unit for inversely converting the level signal into a quantized difference signal output. A quantizer, an adder that adds the quantized difference signal output and the prediction signal, a first frame memory that stores the addition result, and a signal that is predicted in advance as the prediction signal is written. At least one or more third frame memories, and the first and third frames
An image encoding apparatus comprising an adaptive prediction unit that outputs an optimum prediction signal if the image data is stored in the frame memory. 6. A subtractor for taking a difference between an input image signal and a prediction signal, a quantizer for converting the difference signal output into a level signal, and an inverse unit for inversely converting the level signal into a quantized difference signal output. A quantizer, an adder that adds the quantized difference signal output and the prediction signal, a first frame memory that stores the addition result, and an arbitrary block unit in the frame of the input image signal An average value calculation unit for obtaining an average value, an inter-frame difference absolute value sum calculation unit for the calculated average values, and a scene change detection unit for detecting a scene change for each of the arbitrary units according to the magnitude of the difference absolute value sum. A second memory for storing at least one addition result controlled by the scene change detection signal, and image data stored in the first frame memory and the second memory. An image coding apparatus, comprising: an adaptive prediction unit that outputs an optimum prediction signal from the data. 7. The memory according to claim 6, further comprising a memory for writing scene change results in adjacent ranges in the same frame, and performing scene change detection based on the result.
The image encoding device according to the item.