JPH0846976A - Predictive coding device - Google Patents

Abstract

PURPOSE:To provide a predictive coding device capable of preventing deterioration of picture by controlling a quantization width in accordance with the presence/absence of the correlation of pictures within space time. CONSTITUTION:Input and output picture data of a frame memory 11 and output picture data of a frame memory 12 are respectively inputted to a correlation coefficient calculation part 13 and the part 13 divides them into the blocks of the picture elements of 8X8 so as to calculate a correlation coefficient between a present output frame picture and respective frame pictures before and after it by one frame in the area and a correlation coefficient between the lines in horizontal and vertical directions within a present frame. A judgement part 16 decides the quantization width of a quantization part 17 in accordance with a output data quantity from an output buffer memory 24 and furthermore, at the time of judging that the sum sigma of the correlation coefficients is smaller than a threshold value kappa and not-correlated and that the distribution of the frequency spectrums of an information source is uniform based on an orthogonal transforming coefficient from an orthogonal transformation part 15, the part 16 judges the small area of the block of the picture element of 8X8 which is the object is a not-correlated part and makes the quantization width rough.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a predictive coding device,
In particular, the present invention relates to a predictive coding device in a high efficiency coding system for compressing image information.

[0002]

2. Description of the Related Art Conventionally, it has been known to perform predictive coding for image information compression (for example, Japanese Patent Laid-Open No. 3-18).
4487 publication). FIG. 4 shows a block diagram of an example of a conventional predictive coding apparatus that performs such predictive coding. As shown in the figure, the predictive coding apparatus includes a difference unit 31 that takes a difference between image data to be transmitted and image data of one frame before, and an orthogonal transformation unit 3 that orthogonally transforms the difference data of the image.
2, a quantizer 33 for quantizing the orthogonally transformed data,
Inverse quantization unit 34 for restoring quantized data, inverse orthogonal transformation unit 35 for performing inverse transformation of orthogonally transformed data, addition unit 36, frame memory 37, delay unit 38, variable length coding unit 39, output It is composed of a buffer memory 40 and a determination unit 41.

The adding section 36 restores the image by adding the current difference data to the image data of one frame before, and the frame memory 37 stores the restored image data. In addition, the delay unit 38 adjusts the timing of the addition unit 36 that creates the restored image of the previous frame. In addition, the variable length coding unit 39 converts the quantized data into a variable length code. The output buffer memory 40 temporarily holds the encoded data. Further, the determination unit 41 monitors the amount of data accumulated in the output buffer memory 40 and determines whether to change the quantization width of the quantization unit 33.

Next, the operation of this conventional device will be described. The image data to be transmitted is input to the difference unit 31 and is subtracted from the image data of one frame before read from the frame memory 37 to obtain difference data indicating a prediction error. It is supplied to 32 and is orthogonally transformed. The data (transformation coefficient) obtained by the orthogonal transform is quantized by the quantizing unit 33 and then supplied to the variable length coding unit 39, while being supplied to the dequantizing unit 34 and converted into the data before quantization. Restored.

The output data of the inverse quantizing unit 34 is further supplied to the inverse orthogonal transforming unit 35 to be data corresponding to the input difference data of the orthogonal transforming unit 32, and then the adding unit 36 adjusts the time in the delaying unit 38. The original image data is restored by adding the difference data of the previous frame.
The restored image data is delayed by one frame by the frame memory 37 and then input to the difference unit 31 and the delay unit 38.

On the other hand, the variable length coding unit 39 codes (entropy coding) a value having a high appearance frequency in the output data of the quantizing unit 33 after a code having a length shorter than that of a value having a low appearance frequency. As a result, the output variable-length code is generated at an irregular rate in the encoding unit for further increasing the compression rate, so that the output variable-length code is once stored in the output buffer memory 40 in order to keep it at a constant rate.

Here, in order to keep the average of the coding rates constant, the determination unit 41 determines the variable length coded data amount per unit time accumulated in the output buffer memory 40, which increases or decreases depending on the property of the image to be predictively coded. And the quantization width of the quantization unit 33 is variably controlled according to the data amount.

That is, the operation of the judging section 41 will be described. When the encoded data amount increases due to the nature of the input image data at a certain time, and as a result, the data storage amount of the output buffer memory 40 increases, the next input The amount of data to be encoded is reduced by roughening the quantization width of the image data to be encoded. On the contrary, the amount of encoded data decreases due to the nature of the input image data at a certain time, and as a result,
When the amount of data stored in the output buffer memory 40 decreases, the quantization width of the image data to be input next becomes finer and the amount of data to be encoded increases. By such an operation, feedback control is performed so that the amount of data accumulated in the output buffer memory 40 becomes substantially constant.

[0009]

However, in the above-mentioned conventional apparatus, since the quantization width of the quantization unit 33 is controlled from the amount of data stored in the output buffer memory 40 after quantization, the FM image is input to the input image. When there is a non-correlated image area such as noise in the spatiotemporal space, the amount of code in the variable length coding unit 39 is extremely increased, and as a result, a significant image having a correlation is transmitted in the same spacetime. There is a problem that the code amount is relatively reduced and the image quality is deteriorated.

[0010] The present invention has been made in view of the above points,
An object of the present invention is to provide a predictive coding device capable of preventing image quality deterioration by controlling the quantization width depending on the presence or absence of correlation of images in space-time.

[0011]

In order to achieve the above object, the present invention transforms a prediction error of input image data into an orthogonal transform coefficient by an orthogonal transform unit, further quantizes it by a quantizing unit, and then a variable length code. In a predictive coding device that outputs variable-length coded data in an encoding unit through an output buffer memory, memory means for accumulating a plurality of frames of input image data, image data read from the memory means, and input image data, respectively. Based on the correlation coefficient calculation unit that obtains the sum of correlation coefficients between vertical lines, horizontal lines, and frames of image data of a predetermined two-dimensional pixel block, and the amount of data stored in the output buffer memory. And set the quantization width of the quantizer, and at least when the sum of correlation coefficients is smaller than a predetermined threshold, It is obtained by a configuration having a decision section roughening the width.

Further, the judging unit of the present invention judges that the frequency spectrum distribution of the information source is uniform based on the orthogonal transform coefficient, and the sum of the correlation coefficients is smaller than a predetermined threshold value. It is sometimes preferable to coarsen the set quantization width because it is possible to more accurately coarsen the quantization width of only the coded data in the non-correlated image region in the space-time.

[0013]

According to the present invention, when the sum of correlation coefficients between vertical lines, horizontal lines, and frames of image data of a predetermined two-dimensional pixel block is smaller than a threshold value, the image block is excluded. Since the set quantization width is determined to be coarse by determining that it is a correlated portion, it is possible to reduce the amount of encoded data in a significantly uncorrelated image region such as FM noise.

[0014]

Next, an embodiment of the present invention will be described. 1 is a block diagram of an embodiment of a predictive coding apparatus according to the present invention, FIG. 2 is a block diagram of an example of a correlation coefficient calculation unit of FIG. 1,
FIG. 3 shows an operation explanatory diagram of an example of the determination unit in FIG. In this embodiment, as shown in FIG. 1, two frame memories 11 and 12 are connected in cascade, and the frame memories 11 and 12 are connected in series.
Input / output image data of the correlation coefficient calculation unit 13,
Difference unit 14, orthogonal transformation unit 15, determination unit 16, quantization unit 1
7, inverse quantizer 18, inverse orthogonal transformer 19, adder 20,
Frame memory 21, delay unit 22, variable length coding unit 23
And an output buffer memory 24.

That is, in this embodiment, the frame memories 11 and 1 are provided on the input side of the conventional predictive coding apparatus shown in FIG.
2 is provided, and a correlation coefficient calculation unit 13 to which the input / output image data of the frame memories 11 and 12 is input is provided.
The determination unit 41 of the related art is changed to a determination unit 16 that determines the quantization width based on the output of the correlation coefficient calculation unit 13 and the orthogonal transform coefficient.

Next, the operation of this embodiment will be described.
The image data to be transmitted is stored in the frame memory 11 of FIG.
The frame is stored and then read, and further, one frame is stored and read in the frame memory 12. Also,
The input image data and output image data of the frame memory 11 and the output image data of the frame memory 12 are supplied to the correlation coefficient calculation unit 13.

The correlation coefficient calculation unit 13 is configured as shown in FIG. 2, and the output image data of the frame memory 11 is input as the image data of the current frame, and the frame memory 1
The output image data of 2 is input as image data of one frame before (past), and the input image data of the frame memory 11 is input as image data of one frame later (future), and is divided into blocks of 8 × 8 pixels. , The correlation coefficient between the current output frame image in the area and each frame image one frame before and one frame after and the correlation coefficient between horizontal and vertical lines in the current frame are calculated.

That is, as shown in FIG. 2, the correlation coefficient calculation unit 13 includes an inter-frame correlation coefficient calculation unit 131, a horizontal line correlation coefficient calculation unit 132, a vertical line correlation coefficient calculation unit 133, and an inter-frame correlation coefficient calculation unit 133. The correlation coefficient calculation unit 134 and the calculation unit 135 are included in the correlation coefficient calculation unit 132 between horizontal lines. First, the correlation coefficient calculation unit 132 between horizontal lines 8 × 8 shown in Table 1 below.
Extract pixels.

[0019]

[Table 1] Here, in Table 1 above, h _mn indicates the pixel at the intersection of the m-th horizontal pixel row and the n-th vertical pixel row (where 0
≤ m, n ≤ 7). Then, the horizontal line correlation coefficient calculation unit 132 calculates the horizontal line correlation coefficient Rhi in the 8 × 8 pixel block shown in Table 1 based on the following equation.

[0020]

[Equation 1] Similarly, the correlation coefficient Rvj between vertical lines is calculated by the correlation coefficient calculation unit 133 between vertical lines. Further, the correlation coefficient Rf1 between the current frame and the block of 8 × 8 pixels one frame after is calculated by the inter-frame correlation coefficient calculation unit 131 as the correlation coefficient between the block of 8 × 8 pixels one frame before and the current frame. Rf0 is calculated by the inter-frame correlation coefficient calculation unit 134 in the same manner as described above.

The correlation coefficients Rhi, Rvj, and Rfk (k = 1, 2) calculated in this way are
In the calculation unit 135 in the correlation coefficient calculation unit 13 of FIG.
The coefficients α, β, and γ are respectively multiplied and then added to obtain the sum σ of the correlation coefficients represented by the following equation, which is output.

[0022]

[Equation 2] However, in the equation (2), the coefficients α, β and γ are α + β +
There is a relation of γ = 1.

The sum σ of the correlation coefficients is supplied to the judgment unit 16 in FIG. On the other hand, the difference unit 14 of FIG. 1 calculates the difference between the image data of the current frame output from the frame memory 11 and the image data of the previous frame from the frame memory 21, and supplies the difference data to the orthogonal transformation unit 15. , Orthogonal transform. The determination unit 16 receives the sum σ of the correlation coefficients, the orthogonal transformation coefficient from the orthogonal transformation unit 15, and the output data amount from the output buffer memory 24, respectively, and performs the quantization width by the procedure shown in FIG. To decide.

That is, the determination unit 16 first sets a coarse quantization width when the amount of output data from the output buffer memory 24 is large, and finely sets the quantization width when it is small (step 161). Next, the determination unit 16 compares the sum σ of the correlation coefficients with an appropriate threshold value κ for determining the presence or absence of correlation, and determines the frequency of the information source from the distribution of the orthogonal coefficients from the orthogonal transformation unit 15. It is determined whether the spectral distribution is uniform (step 162).

For example, if the judgment unit 16 judges in step 162 that the sum σ of correlation coefficients is smaller than the threshold value κ and there is no correlation, and that the frequency spectrum distribution of the information source is uniform, then it is regarded as a target. It is determined that the small area of the 8 × 8 pixel block is a non-correlated portion, and the quantization width is made coarse (step 164).

On the other hand, when the judgment unit 16 detects in step 162 that the sum σ of correlation coefficients is equal to or larger than the threshold value κ and there is a correlation, or when the frequency spectrum distribution of the information source is not uniform, , The quantization width is not changed (step 163). The determination result generated by the determination unit 16 by such an algorithm is supplied to the quantization unit 17 in FIG. 1 to control the quantization width.

The quantizing unit 17 quantizes the output orthogonal transform coefficient of the orthogonal transforming unit 15 with a quantization width according to the determination result from the determining unit 16, and the obtained signal is dequantized by the inverse quantizing unit 18 and the variable length. It is supplied to the encoding unit 23. The signal inversely quantized by the inverse quantizer 18 and inversely orthogonally transformed by the inverse orthogonal transformer 19 is supplied to the adder 20 and is input via the delay unit 22 from the frame memory 21 one frame before. Signal is added to restore the original image data and stored in the frame memory 21. The image data read from the frame memory 21 is input to the difference unit 14 and the delay unit 22.

Further, the variable length coded by the variable length coding unit 23 in the same manner as in the conventional case is temporarily stored in the output buffer memory 24 and then output at a constant rate.
Here, in this embodiment, the determination unit 16 causes the quantization unit 17 to operate.
Since the quantization width of is controlled by the above-mentioned algorithm, the quantization width of the image data of the uncorrelated portion in the space-time is made coarse, and the amount of data that has been variable-length coded is smaller than in the past. As a result, the amount of data stored in the output buffer memory 24 decreases.

Therefore, in the present embodiment, the image data of the part having the same spatiotemporal correlation has a relatively fine quantization width because the increase in the data amount of the image data having no correlation is suppressed as compared with the conventional case. Is quantized, and image quality deterioration is suppressed.

In the above embodiment, the determination unit 16 determines whether or not the frequency spectrum of the information source is uniform on the basis of the orthogonal transform coefficient, thereby more accurately determining whether or not the image is significantly uncorrelated in space and time. However, in principle, it is possible to determine whether or not to make the quantization width fine only by comparing the sum σ of the correlation coefficients and the threshold value κ.

[0031]

As described above, according to the present invention,
An image of a significantly uncorrelated region in space-time (FM noise, etc.)
Since the coded data amount of can be reduced compared to the conventional one, the coded data amount of a significant image area having a correlation in the same spatiotemporal space can be increased, and thus, a significant correlation The image quality of the image can be improved as compared with the conventional one.

[Brief description of drawings]

FIG. 1 is a block diagram of one embodiment of the present invention.

FIG. 2 is a block diagram of an example of a correlation coefficient calculation unit in FIG.

FIG. 3 is an operation explanatory diagram of an example of a determination unit in FIG.

FIG. 4 is a block diagram of an example of a conventional device.

[Explanation of symbols]

 11, 12, 21 Frame memory 13 Correlation coefficient calculation unit 14 Difference unit 15 Orthogonal transformation unit 16 Judgment unit 17 Quantization unit 17 23 Variable length coding unit 24 Output buffer memory 131, 134 Interframe correlation coefficient calculation unit 132 Horizontal Inter-line correlation coefficient calculation unit 133 Vertical inter-line correlation coefficient calculation unit 135 Operation unit

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Takeo Tsutsui 2-2-1, Jinnan, Shibuya-ku, Tokyo Inside the Japan Broadcasting Corporation Broadcasting Center (72) Inventor, Nao Miyoshi 2-2-1, Jinnan, Shibuya-ku, Tokyo Kyodo Broadcast Center

Claims (3)

[Claims] 1. A prediction error of input image data is converted into an orthogonal transformation coefficient by an orthogonal transformation unit, further quantized by a quantization unit, and then variable length encoded data is output by a variable length encoding unit through an output buffer memory. In the predictive coding device, an image of a predetermined two-dimensional pixel block is formed by using memory means for accumulating the input image data in a plurality of frames, and image data read from the memory means and the input image data. A correlation coefficient calculation unit that calculates the sum of correlation coefficients between vertical lines of data, horizontal lines, and frames, and sets the quantization width of the quantization unit based on the amount of data stored in the output buffer memory. And at least when the sum of the correlation coefficients is smaller than a predetermined threshold value, the determination unit that coarsens the set quantization width. And a predictive coding device. 2. The determining unit determines that the frequency spectrum distribution of the information source is uniform based on the orthogonal transform coefficient, and the sum of the correlation coefficients is smaller than a predetermined threshold value. The predictive coding apparatus according to claim 1, wherein the set quantization width is made coarse at times. 3. The correlation coefficient calculation unit uses the image data read from the memory means and the input image data, respectively, to determine a phase between horizontal lines of image data of a predetermined two-dimensional pixel block. A horizontal line correlation coefficient calculation unit that calculates a relation number, a vertical line correlation coefficient calculation unit that calculates a correlation coefficient between vertical lines of image data of the two-dimensional pixel block, and a two-dimensional pixel block An inter-frame correlation coefficient calculation unit that calculates a correlation coefficient between the current frame of the image data and each frame before and after one frame, the horizontal line correlation coefficient calculation unit, and the vertical line correlation coefficient calculation unit 2. The predictive code according to claim 1, further comprising: an arithmetic unit that multiplies each output correlation coefficient of the interframe correlation coefficient calculation unit by a constant and then adds the product to output the sum of the correlation coefficients. Conversion apparatus.