JPH0360293A - Method and apparatus for compressing video signal and method and apparatus for expanding video signal - Google Patents

Abstract

PURPOSE:To trace a steep change in horizontal and vertical directions and to reduce gradient overload noise by selecting three prediction modes of horizontal linear, vertical linear and 2-dimension so as to control a gain switching circuit. CONSTITUTION:Three data, that is, a data resulting from a pre-data being an output of an adder 4 through a delay circuit 10a for timing adjustment, a nearly 1H delay data through a delay circuit 6a and a nearly (1H-1) picture element delay data being an output of a delay circuit 5a are inputted to a prediction mode deciding circuit 11a. The prediction mode deciding circuit 11a takes the absolute value of the difference of the three data in the horizontal and vertical directions based on the quantity and switching signals of three prediction modes of horizontal linear, vertical linear and 2-dimension are outputted to control gain changeover circuits 7a, 7b. Thus, the tracking to a rapid change in the vertical and horizontal directions is enhanced and the gradation overload noise is reduced.

Description

Detailed Description of the Invention [Field of Industrial Application] The present invention relates to a method and apparatus for compressing a video signal, and a method and apparatus for expanding a video signal, and in particular, a method and apparatus for compressing a video signal using differential pulse modulation, and a method and apparatus for expanding a video signal. The present invention relates to a method and apparatus for performing the same.

[Conventional technology]

2. Description of the Related Art Differential pulse modulation (DPCM) is widely used, in which data is compressed by quantizing a video signal and obtaining a difference between a signal at a pixel in the vicinity of the pixel undergoing quantization. If a video signal is compressed using this DPCM method, video signals with high redundancy, that is, high line correlation or frame correlation, can be represented with fewer bits than when quantized without being compressed. This is useful for sending and recording. A video signal compressed by the DPCM method can be expanded using a predictor used during compression.

FIG. 8 is a block diagram of a circuit that compresses and expands a video signal using the widely used DPCM method.

In FIG. 8, the image signal to be encoded is sampled and added to point A. The predicted value of the next sampled value is added to point C as the output of the prediction circuit included in the feedback loop, and a difference signal from the actual sampled value is obtained at point B. Then, quantization and appropriate encoding are performed, and the signal is sent out as an encoded signal via point D. On the other hand, the output signal is guided to a feedback loop, and by summing it with the predicted value that was previously subtracted, an approximate value of the signal at point A is obtained at point E, including only the quantization error. It turns out.

Here, as the prediction method shown in FIG. 8, one of the following three functions is used.

When performing one-dimensional prediction in the horizontal direction x-a When performing one-dimensional prediction in the vertical direction x-c When performing two-dimensional prediction x-(
a+c)/2 or x-(a+d)/2
Note that when performing quantization, visual characteristics are taken into consideration and quantization is performed as a logarithmic difference.

[Problem that the invention attempts to solve]

In the conventional video signal compression and expansion method described above, when horizontal one-dimensional prediction is performed as a prediction method, it is impossible to follow sudden changes in the horizontal direction, that is, the predicted value and the actual value may differ greatly. There is also a problem that when vertical one-dimensional prediction is performed, it is not possible to follow sudden changes in the vertical direction, and two-dimensional prediction has the problem that there is a lot of gradient overload noise in the reproduced video. .

SUMMARY OF THE INVENTION An object of the present invention is to provide a video signal compression method and apparatus, as well as an expansion method and apparatus, which improve tracking of steep changes in the vertical and horizontal directions and reduce gradient overload noise.

[Means and operations for solving the problem] In order to achieve the above object, the video signal compression method and device and the decompression method and device of the present invention perform vertical one-dimensional prediction and horizontal one-dimensional prediction as prediction methods. Detecting the correlation between the video signal and the horizontal direction, vertical direction, or both,
Depending on the correlation, output either a vertical one-dimensional predicted level, a horizontal one-dimensional predicted level, or a mixed level of a vertical one-dimensional predicted level and a horizontal one-dimensional predicted level as a two-dimensional predicted level. do. Alternatively, the mixture may be mixed by continuously changing the mixing ratio of the horizontal one-dimensional level and the vertical one-dimensional level according to the above correlation.

In other words, one-dimensional prediction in the horizontal direction and (-) are difficult to follow when there is a sudden change in the horizontal direction (such as a vertical line), and the image becomes unclear, but when there is a sudden change in the vertical direction. (for horizontal lines, etc.), the image can be displayed clearly because the resolution ability of the horizontal line is directly reflected.On the other hand, one-dimensional prediction in the vertical direction is difficult to predict in the horizontal direction for the same reason. It is possible to clearly display sudden changes in

Using this feature, we use the upper pixel "X", the pixel "a", and the upper left pixel "d" shown in FIG. If the difference is within a predetermined value, two-dimensional prediction is performed, and if it exceeds a predetermined value, one-dimensional prediction is performed in the horizontal or vertical direction depending on the sign of the difference.In other words, h
h-vv≧TH・・・・・・・・・・・・・・・(1)
When , one-dimensional prediction is made in the vertical direction, hh-vv≦-TH・・・・・・・・・・・・・・・(
'2) One-dimensional child in the horizontal direction? /91L (TH is the reference value), otherwise two-dimensional prediction is performed. Or the horizontal and vertical predicted values may be mixed in continuous ratios. Alternatively, the predetermined value is set to two levels, and the prediction function xa is used.

You may also do so. In this case, the followability of the compressed signal to the original signal becomes even better.

Furthermore, it is also possible to determine the correlation using only the level difference hh, and the prediction function is switched according to the following conditions.

hh≦THI ・・・・・・・・・・・・
...At the time of (3), one-dimensional prediction in the horizontal direction is performed, and THI<hh<TH2...
When (4), perform two-dimensional prediction, TH2≦hh ・・・・・・・・・・・・
...When (5) is performed, one-dimensional prediction in the vertical direction is performed (THI,
TH2 is a reference value).

This is based on the fact that statistically, when hh is large, the correlation in the vertical direction is high, and when hh is small, the correlation in the horizontal direction is high. In this case, since there is only one condition to be determined, the circuit configuration can be simplified and the response speed can be increased.

That is, video signal compression stores the level of the sampled video signal, calculates the difference between the predicted level determined by the stored level and the current video signal, obtains a video difference signal, quantizes it, and compresses the video signal. In the method, a second pixel N lines before the first pixel constituting the current video signal, a third pixel N lines before the first pixel, and a third pixel N lines before the first pixel and M(M , N is 1
(integer greater than or equal to) pixel) The video signal level of each of the fourth pixel in front of the pixel is stored, the vertical correlation is detected from the video signal level of the third pixel and the S4 pixel, and the video signal level of the second pixel and the fourth pixel is A horizontal correlation is detected from the level, and a first predicted level based on the video signal level of the second pixel stored according to the detected degree of one or both of the vertical correlation and horizontal correlation, and a first predicted level of the stored third pixel. One of a second prediction level based on the video signal level and a third prediction level obtained by mixing the second and third prediction levels at a predetermined ratio.
There is provided a video signal compression method characterized in that the predicted level is obtained by selecting one of the predicted levels.

Further, in a video signal expansion method, the video signal is expanded by dequantizing a quantized and compressed video signal, storing the level, and calculating the sum of the current video signal and a predicted level determined by the stored level, A second pixel N lines before the first pixel that constitutes the current video signal, a third pixel N lines before the first pixel, and a third pixel N lines before the first pixel and M (M, N
is an integer greater than or equal to 1) The video signal level of each of the fourth pixel before the pixel is stored, the vertical correlation is detected from the video signal levels of the third pixel and the fourth pixel, and the vertical correlation between the second pixel and the fourth pixel is detected.
A horizontal correlation is detected from the video signal level of the pixel, and the first one is based on the video signal level of the second pixel, which is stored according to the detected degree of one or both of the vertical correlation and the horizontal correlation.
, a second prediction level based on the stored video signal level of the third pixel, and a third prediction level obtained by mixing the second and third prediction levels at a predetermined ratio. A video signal decompression method is provided, characterized in that the predicted level is obtained by selecting the predicted level.

Furthermore, as shown in FIG. 4, hh is created by the absolute value of the data difference between pixel "C#" two lines before and pixel "f" one pixel before it, and Pixel “e#
A is created from the absolute value of the difference between the levels of the pixel g'' and the pixel g'' one line before it, and the measurement function is switched as follows.

When hh-vv≧TH or hh-hh - (8), one-dimensional prediction is performed in the vertical direction, and when hh-vv≦-TH (7), one-dimensional prediction is performed in the horizontal direction. In other cases, two-dimensional prediction is performed.

This is effective for diagonal lines for which the two methods described above have little improvement. For example, in the case shown in Figure 5, if there is no hh-hh condition, pixel X will be
The prediction is made in one dimension in the vertical direction, resulting in a large error.

In addition, in the case shown in FIG. 6, if there is no vv-mama condition, the pixel X will be predicted in one dimension in the horizontal direction, resulting in a large error. However, if this method is used, the diagonal line will be predicted two-dimensionally, and the tracking performance for the original signal after compression or expansion will be improved.

〔Example〕

Examples will be described with reference to the drawings. FIG. 1 is a block diagram of a video signal compression apparatus using the video signal compression method of the first embodiment of the present invention.

In FIG. 1, a sampled 8-bit video signal, that is, image data, is input to a subtracter 1, and is subtracted by prediction data generated by a prediction circuit 13a. The output data is compressed into 4-bit data using the converted conversion table. Further, the output data is inputted to the inverse quantizer 3a, returned to 8-bit difference data, and inputted to the adder 4. The other input of the adder 4 receives, through the switching circuit L2a, either the predicted data created one pixel before or the 8-bit original data that is periodically inserted into the compressed transmission data. prefix data is created.

The prefix data is input to the gain switching circuit 7a, and the gain is switched according to the output of the prediction mode determination circuit 11a, and then output to the adder 9. On the other hand, the prefix data that is the output of the adder 4 is passed through a delay circuit 5a with a delay amount of about IH minus 1 pixel and a delay circuit 6a with a delay amount of about 1 pixel, and the prefix data is delayed by about IH. Then, the signal is input to the gain switching circuit 7b, and the gain is switched according to the output of the prediction mode determination circuit 11a, and then output to the adder 9. The output of the adder 9 is input to the subtraction side of the subtracter 1 as predicted data.

The prediction mode determination circuit Ha divides the prefix data that is the output of the adder 4 into the data that has passed through the delay circuit 10a for timing adjustment, the approximately IH delayed data that has passed through the delay circuit 6a, and the output of the delay circuit 5a. Three data of approximately IH minus 1 pixel delay data are input manually, and the absolute value of the difference between the horizontal and vertical directions of the three data is taken, and depending on the magnitude, the above formula (L), (2) Due to the conditions of
It outputs switching signals for three prediction modes: horizontal one-dimensional, vertical one-dimensional, and two-dimensional to control gain switching circuits 7a and 7b. Alternatively, it is determined based on the conditions of equations (3), (4), and (5), and a switching signal is output. In the case of horizontal one-dimensional prediction, the gain of the gain switching circuit 7b is set to one flash, and the gain of the gain switching circuit 7a is controlled to be switched to one.

In the case of vertical one-dimensional prediction, control is performed so that the gain of the gain switching circuit 7a is set to 1, and the gain of the gain switching circuit 7b is switched to 1. Further, it is determined that the prediction is two-dimensional prediction, and the prediction is controlled to be switched. Assuming the dynamic range is 8 bits (0 to 255), the TH value of Naotake (1) and (2) is preferably around 30.

Also, depending on the conditions of formulas (3), (4), and (5),
When the prediction mode is determined, THI is set to 6 to 7. TH2
When set to around 24 to 25, an output that followed the original signal well was obtained.

Next, a video signal decompression device using the video signal decompression method according to the second embodiment of the present invention will be described with reference to the block diagram of FIG. 2.

In FIG. 2, the encoded compressed video signal, that is, image data compressed to 4 bits, is input to the inverse quantizer 3b, and is converted to 8-bit data according to a predetermined conversion table. It is input to the adder 22. The other input of the adder 22 is connected to the input data switching circuit 1.
2b, that is, the predicted value data that is the output of the adder 29, or the 8-bit original data that is inserted periodically (for example, once in IH) between the input data compressed to 4 bits per pixel. The selected one is input. The output of the adder 22 becomes encoded data and is output. The output is also fed back to generate predicted value data. First, the output data of the adder 22 is input to the gain switching circuit 7d.

The gain can be selected, and the selection is controlled by the output of the prediction mode determination circuit 11b. The output of the gain switching circuit 7d is input to the adder 29.

Further, the output of the adder 22 passes through a delay circuit 5b having a delay amount of about IH minus 1 pixel and a delay circuit 6b having a delay amount of about 1 pixel.
It is manually operated by c. Also, like the gain switching circuit 7d, it is controlled by the output of the prediction mode determination circuit 11b. The output of the gain switching circuit 7d is input to the adder 29. Adder 29
The output of is input to input data switching circuit 1 as previous value prediction data.
2b. The prediction mode determination circuit flb divides the output data of the adder 22, that is, the decoded data, into the previous value data that has passed through the delay circuit 10b for timing alignment, and the data before IH minus one pixel that has passed through the delay circuit 5b. , the absolute value of the difference value in the horizontal and vertical directions is calculated manually using the 3 data of the data before IH that has passed through the delay circuit 6b, and according to the magnitude, the horizontal 1-dimensional, vertical 1-dimensional, and 2-dimensional
The gain switching circuit 7
c, 7d. That is, formula (1), (2) or (
3) Determine the conditions of (4) and (5) and create a switching signal.

Control in each prediction mode will be explained below.

If horizontal one-dimensional prediction is determined, a control signal is output to switch the gain of gain switching circuit 7c to 1 and the gain of gain switching circuit 7d to 1, and if vertical one-dimensional prediction is determined, gain switching circuit 7c is output. 7C gain to 1, gain switching circuit 7d
A control signal is output to switch the gain of 1 to 1, and a two-dimensional prediction and determination control signal is output. In this way, the three prediction modes are switched, and image data is decoded accordingly.

The TH value of Naotake (1) and (2) is preferably around 30 when the dynamic range is set to 8 bits (0 to 255).

Also, depending on the conditions of formulas (3), (4), and (5),
If the prediction mode is determined, THI is 6 to 7, TH2
When set to around 24 to 25, an output that followed the original signal well was obtained.

Next, a video signal compression apparatus using the video signal compression method according to the third embodiment of the present invention will be described with reference to the block diagram of FIG.

The video signal compression device shown in FIG. 7 has the same basic configuration as the first embodiment, but the prediction mode determination circuit lie is configured to determine the prediction mode to determine the conditions of equations (6) and (7). It is provided in place of the determination circuit 11a, and furthermore, in order to determine the conditions of equations (6) and (7), a delay circuit is provided in order to obtain the pixels "C, e, "f", and g' shown in FIG. circuit 14 and delay circuit], 5a, 15b.

15c is provided. The delay circuit 14 has a delay amount of IH, and the delay circuits 15a, 15b, and 15c have a delay amount of one pixel. Therefore, the prediction mode determination circuit 11c can determine the conditions of equations (6) and (7) and send switching signals to the gain switching circuits 7a and 7b.

Further, as a fourth embodiment, a prediction circuit 13c shown in FIG.
If applied to the video signal expansion device of FIG. 2 in place of the prediction circuit 13b, the following will be obtained: The gain switching circuits 7c and 7d are controlled by the switching signals whose correlations are determined according to the conditions of equations (8) and (7). A signal expansion device is provided.

In addition, when the prediction mode was determined according to the conditions of equations (8) and (7), good followability was shown around 30 within the 8-bit (0 to 255) range. In this case hh and h
h, vv, and m may be approximately the same.

〔Effect of the invention〕

As explained in detail above, the video signal compression method and decompression method of the present invention can reduce gradient overload noise in video with the same amount of information and provide video closer to the original image than the conventional DPCM method. can. Alternatively, when providing the same level of video, it is possible to reduce the amount of information and achieve more efficient compression.

[Brief explanation of drawings]

FIG. 1 is a block diagram of a video signal compression device using the video signal compression method of the first embodiment of the present invention, and FIG. 2 is a block diagram of a video signal compression device using the video signal expansion method of the second embodiment of the present invention. A block diagram of a signal expansion device, FIGS. 3 to 6 are diagrams showing the relationship between pixels and video signals for explaining the present invention, and FIG. 7 is a block diagram of a video signal compression device according to a third embodiment. FIG. 8 is a block diagram including the predictor of the fourth embodiment, and a block diagram of a conventional video signal compression device and decompression device. 1... Subtractor, 2... Quantizer, 3a, 3b...
Inverse quantizer, 4. 9.22.29--Adder, 5
a, 5 b, 6 a. 6 b, loa, 10b, 14.15a, 15
b, 15a--Delay circuit, 7a. 7b, 7c, 7d--Gain switching circuit, lla, llb
--Prediction mode determination circuit, 13a, 13b, 1
3c...Prediction circuit. Inventor Takayuki Sugawara Katsuyuki

Claims (6)

[Claims] (1) A video image that stores the level of a sampled video signal, takes the difference between the predicted level determined by the stored level and the current video signal, obtains a video difference signal, quantizes it, and compresses the video signal. In the signal compression method, a second pixel N lines before the first pixel constituting the current video signal, a third pixel N lines before the first pixel and M pixels before the first pixel, and a third pixel N lines before the first pixel and M pixels (M and N are 1
(an integer greater than or equal to) pixel) The video signal level of each of the fourth pixel in front of the pixel is stored, the vertical correlation is detected from the video signal levels of the third pixel and the fourth pixel, and the video signal level of the second pixel and the fourth pixel is detected. A horizontal correlation is detected from the signal level, and a first predicted level based on the video signal level of the second pixel is stored according to the detected degree of one or both of the vertical correlation and the horizontal correlation, and the third pixel is stored. The predicted level is obtained by selecting one of a second predicted level based on the video signal level and a third predicted level obtained by mixing the second and third predicted levels at a predetermined ratio. A video signal compression method. (2) A video signal expansion method that dequantizes a quantized and compressed video signal, stores the level, and expands the video signal by adding the predicted level determined by the stored level and the current video signal. A second pixel N lines before the first pixel constituting the current video signal, a third pixel M pixels before the first pixel, and a third pixel N lines before the first pixel and M(M , N is an integer greater than or equal to 1) The video signal level of each of the fourth pixel before the pixel is stored, the vertical correlation is detected from the video signal levels of the third pixel and the fourth pixel, and the vertical correlation between the second pixel and the fourth pixel is detected. A horizontal correlation is detected from the video signal level of the four pixels, and a first predicted level based on the video signal level of the second pixel is stored according to the degree of one or both of the detected vertical correlation and horizontal correlation, and The predicted level is set by selecting one of a second predicted level based on the video signal level of the third pixel and a third predicted level obtained by mixing the second and third predicted levels at a predetermined ratio. A video signal expansion method characterized in that: (3) A video image that stores the level of the sampled video signal, takes the difference between the predicted level determined by the stored level and the current video signal, obtains a video difference signal, quantizes it, and compresses the video signal. In the signal compression method, a second pixel N lines before the first pixel constituting the current video signal, a third pixel N lines before the first pixel and M pixels before the first pixel, and a third pixel N lines before the first pixel and M pixels the previous fourth pixel, the fifth pixel N+n lines before the first pixel, the sixth pixel N+n lines before the first pixel and M pixels before, and the seventh pixel M+m pixels before the first pixel. and an eighth pixel that is N lines before the first pixel and M+m (M, N, m, n is an integer of 1 or more) pixels before the third pixel, 1st from 4 pixel video signal level
the absolute value of the second difference from the video signal level of the second pixel and the fourth pixel, and the absolute value of the third difference from the video signal level of the fifth pixel and the sixth pixel. The absolute value of the fourth difference is detected from the video signal level of the seventh pixel and the eighth pixel, and the absolute value of the first difference is subtracted from the detected absolute value of the second difference. When the first condition is that the value is larger than a predetermined value and the absolute value of the second difference and the absolute value of the third difference are approximately equal, a first predicted level based on the stored video signal level of the second pixel. is determined by the stored video signal level of the third pixel when the subtracted value is smaller than a predetermined value and the absolute value of the first difference and the absolute value of the fourth difference are approximately equal. When the second prediction level is other than the first and second conditions, selecting a third prediction level obtained by mixing the second and third prediction levels at a predetermined ratio to obtain the prediction level. A video signal compression method characterized by: (4) A video signal expansion method that dequantizes a quantized and compressed video signal, stores the level, and expands the video signal by adding the predicted level determined by the stored level and the current video signal. A second pixel N lines before the first pixel constituting the current video signal, a third pixel N lines before the first pixel and M pixels before the first pixel, and a third pixel N lines before the first pixel and M pixels before the first pixel. 4 pixels, a fifth pixel that is N+n lines before the first pixel, a sixth pixel that is N+n lines and M pixels before the first pixel, a seventh pixel that is M+m pixels before the first pixel, and the third pixel. The respective video signal levels of the 8th pixel N lines before the 1 pixel and M+m (M, N, m, n are integers of 1 or more) pixels before are stored, and the video signal levels of the 3rd pixel and the 4th pixel are stored. 1st from the video signal level
the absolute value of the second difference from the video signal level of the second pixel and the fourth pixel, and the absolute value of the third difference from the video signal level of the fifth pixel and the sixth pixel. The absolute value of the fourth difference is detected from the video signal level of the seventh pixel and the eighth pixel, and the absolute value of the first difference is subtracted from the detected absolute value of the second difference. When the first condition is that the value is larger than a predetermined value and the absolute value of the second difference and the absolute value of the third difference are approximately equal, a first predicted level based on the stored video signal level of the second pixel. is determined by the stored video signal level of the third pixel when the subtracted value is smaller than a predetermined value and the absolute value of the first difference and the absolute value of the fourth difference are approximately equal. When the second prediction level is other than the first and second conditions, selecting a third prediction level obtained by mixing the second and third prediction levels at a predetermined ratio to obtain the prediction level. A video signal expansion method characterized by: (5) A video image that stores the level of the sampled video signal, takes the difference between the predicted level determined by the stored level and the current video signal, obtains a video difference signal, quantizes it, and compresses the video signal. In the signal compression device, a second pixel N lines before the first pixel constituting the current video signal, a third pixel N lines before the first pixel, and a third pixel N lines before the first pixel and M pixels (M and N are 1
means for storing the video signal level of each of the fourth pixel in front of the pixel (an integer greater than or equal to) pixel; and detecting a vertical correlation from the video signal levels of the third pixel and the fourth pixel; means for detecting a horizontal correlation from the video signal level of the second pixel; a first predicted level based on the video signal level of the second pixel stored according to the detected degree of one or both of the vertical correlation and the horizontal correlation; The predicted level is determined by selecting one of a second predicted level based on the video signal level of the third pixel and a third predicted level obtained by mixing the second and third predicted levels at a predetermined ratio. A video signal compression device characterized by having means for obtaining. (6) A video signal expansion device that dequantizes a quantized and compressed video signal, stores the level, and expands the video signal by calculating the sum of the predicted level determined by the stored level and the current video signal. , a second pixel N lines before the first pixel constituting the current video signal, a third pixel N lines before the first pixel, and a third pixel N lines before the first pixel and M(M, N is 1
means for storing the video signal level of each of the fourth pixel in front of the pixel (an integer greater than or equal to) pixel; and detecting a vertical correlation from the video signal levels of the third pixel and the fourth pixel; means for detecting a horizontal correlation from the video signal level of the second pixel; a first predicted level based on the video signal level of the second pixel stored according to the detected degree of one or both of the vertical correlation and the horizontal correlation; The predicted level is set by selecting one of a second predicted level based on the video signal level of the third pixel and a third predicted level obtained by mixing the second and third predicted levels at a predetermined ratio. 1. A video signal decompressing device, comprising means for decompressing a video signal.