JPH0591338A - Video informaiton volume compressing and/or expanding device - Google Patents

Abstract

PURPOSE:To suppress the noise based on quantization noise. CONSTITUTION:This video information volume compressing or information volume compressing and expanding device consists of a main encoding means AA which is provided with a first encoding means AA1 which outputs a first encoded signal 10a obtained by encoding of 8X8 DCT blocks 1a obtained by division of a video signal and a second encoding means AA2 which outputs a second encoded signal 13a obtained by encoding of 4X4 subblocks, a control means BB which detects a high-band space frequency component of DCT blocks la and outputs a control signal 20a based on this component, and a means which selects and multiplexes first and second encoded signals based on the control signal 20a.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION The present invention relates to video information amount compression and / or
Alternatively, the present invention relates to an information amount expansion device.

[0002]

2. Description of the Related Art FIG. 11 is a block diagram of a conventional still image compression apparatus, and FIG. 12 is a diagram for explaining the operation of a zigzag scanner.

Referring to FIG. 11, JPEG (Joint Photog) is an organization that promotes international standardization of color still image coding.
In order to explain the conventional still image compression device (see P158 Vol44 No2 1990 of the Institute of Television Engineers of Japan) proposed by the raphic Expert Group), first, 8 pixel vertical and 8 horizontal digital pixel data are extracted from one screen. The obtained DCT block is D
It is supplied to the CT converter 1 (8 × 8 DCT), and the 8 × 8 transform coefficient obtained by performing the two-dimensional DCT conversion here is supplied to the linear quantizer 2. Here, DCT is an abbreviation for Discrete Cosine Transform, which is a type of orthogonal transform coding.

Then, the reference quantization matrix supplied from the quantization matrix generator 3 (quantization matrix) is multiplied by the scaling factor S in the multiplier 4 to perform linear quantization using the obtained quantization matrix. The DC conversion coefficient 2a and the AC conversion coefficient 2b thus obtained are supplied to the one-dimensional predictor 5 and the zigzag scanner 7, respectively.

The one-dimensional predictor 5 compresses the information amount of the DC conversion coefficient 2a by obtaining the difference value between the continuous DC conversion coefficients 2a and performing one-dimensional prediction, and supplies the compressed information to the first Huffman encoder 6. To do. Then, the first Huffman encoder 6 generates a Huffman code that allocates a smaller number of bits in the descending order of occurrence probability of an event, and supplies the Huffman code to one input of the multiplexer 9. The Huffman code is a type of entropy code because it is based on the probability of occurrence of an event. Further, the occurrence probability of the event may be obtained by using the result measured in advance, and it is not necessary to obtain it at the time of encoding.

On the other hand, the zigzag scanner 7 scans the AC conversion coefficient 2b from the low frequency component to the high frequency component as shown in FIG. 12, and supplies this to the second Huffman encoder 8. Then, the second Huffman encoder 8 similarly generates a Huffman code based on the run length of the transform coefficient whose value is “0” and the value of the transform coefficient which is not “0”, and the other Huffman code of the multiplexer 9 is Supply to the input.

Then, the multiplexer 9 outputs a multiplexed output signal 9a obtained by multiplexing the respective inputs to a transmission line (not shown).

Since the above-mentioned conventional technique uses the Huffman code, the amount of information contained in the multiplexed output signal 9a varies with respect to a constant amount of input information. Therefore, the scaling factor S is used to control, for example, the information amount of the multiplexed output signal 9a per field to be held constant.

[0009]

However, in the above-mentioned conventional technique, since the quantization noise described later is generated in the linear quantizer 2, there is a problem that there is a DCT block in which the image quality is significantly deteriorated. Hereinafter, in order to facilitate understanding of the problem, a one-dimensional DCT will be described as an example.

FIG. 13 is a diagram for explaining image quality deterioration due to compression / expansion using DCT due to quantization noise, and FIG. 14 is a diagram for explaining a quantization matrix. Problems to be solved by the present invention will be specifically described below with reference to the drawings.

Here, FIG. 13 having a periodic change.
One-dimensional digital waveform data string shown in (A) and amplitude 1
1 shown in (C) of FIG. 1 which is composed of 1 and a flat portion of amplitude 0.
Consider a three-dimensional digital waveform data string. In addition, reverse DC
The T transform is defined by Equation 1 below. In addition, FIG.
The vertical axis in (A) to (D) represents the level of the video signal.

[0012]

[Equation 1] Then, the above-mentioned data string is subjected to a one-dimensional DCT transform defined by the following Equation 2 to quantize the transform coefficient D (7) in a coarse quantizing step, and then to dequantize and inverse DCT transform. The waveform data strings obtained by the above are shown in FIG. 7B and FIG.

[0013]

[Equation 2] Now, here, the portions shown by the dotted lines in FIGS. 6B and 6D are noises based on the quantization noise generated by the quantization in the coarse quantization step, but considering the visual characteristics,
In the waveform shown in FIG. 7B, small noise is added to a place where there is a large change in the video signal level, so the differential sensitivity (“Digital signal processing of image”, Nikkan Kogyo Shimbun (1987) P2
3) is low, it is difficult to detect noise. On the other hand, the waveform shown in FIG. 3D has high differential sensitivity. Therefore, noise is easily detected, and the noise appearing on the flat portion becomes a problem.

Although the case of one dimension has been described above,
The same applies to the two-dimensional case. The quantization matrix shown in FIG. 14 has a value that increases as the horizontal and vertical frequencies increase in the diagonally lower right direction. That is, the higher the transform coefficient, the coarser the quantization step in linear quantization. This is because the amount of information is compressed by utilizing the fact that the luminosity is lowered in the portion where the spatial frequency is high and the luminosity is lowered with respect to the waveform in the oblique direction.

However, even in the case of two dimensions, FIG.
A waveform composed of a flat portion and a contour portion as in (C) has a high frequency component, and such a waveform has a two-dimensional DC.
When the transform coefficient obtained by applying the T transform is quantized using the above quantization matrix, quantization noise is mixed in the quantized transform coefficient of higher order (lower right diagonal). Then, when this is inversely quantized and subjected to inverse two-dimensional DCT conversion, noise based on quantization noise is generated, which appears as a noticeable disturbance particularly in a flat portion.

Therefore, the present invention has been made in view of the above-mentioned problems.
For a DCT block having a flat portion and a contour portion, this DCT block is divided and subjected to DCT transformation.

[0017]

The present invention provides the following constitution in order to solve the above-mentioned problems.

The input pixel block obtained by dividing the video signal is subjected to DCT conversion, quantized, and encoded to output a first encoded signal, and the input pixel block is divided. Each of the plurality of sub blocks obtained by
Main encoding means having a second encoding means for outputting a second encoded signal obtained by converting, quantizing, and encoding;
A control signal for selecting the first coded signal and the second coded signal for each input pixel block based on the signal obtained by detecting the high spatial frequency component of the input pixel block. And a selection signal obtained by selecting one of the first coded signal and the second coded signal based on the control signal, based on the control signal. And a video information amount compression or information amount compression / expansion device.

A video information amount expanding device or an information amount compressing / expanding device for expanding the information amount compressed signal according to claim 1, wherein the first encoding is performed based on the decoded discrimination information. A video information amount expansion or information amount compression / expansion device, characterized by decoding the information amount compression signal having a signal and the second coded signal.

[0020]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a block diagram for explaining a main part of a video information amount compression apparatus of the first embodiment, and FIG. 2 is a numerical value for explaining a main part of a video information amount compression apparatus of the first embodiment. FIG. 3 shows an example, FIG. 3 is a circuit diagram for explaining an example of the first and second area discriminators, and FIG. 4 is a block for explaining the main part of the video information amount expansion device of the first embodiment. Figure, Figure 5 is the first
FIG. 6 is a diagram showing a numerical example for comparing the embodiment with a conventional technique, FIG. 6 is a table showing an example of Huffman coding, and FIG.
9 is a block diagram for explaining a main part of the video information amount compression apparatus of the second embodiment, FIG. 8 is a signal flow graph for explaining a high-speed algorithm of one-dimensional DCT conversion, and FIG.
Is a block diagram for explaining the first and second one-dimensional DCT converters, and FIG. 10 is a block diagram for explaining the main part of the video information amount expansion apparatus of the second embodiment. The first embodiment will be described below with reference to the drawings.

(First Embodiment) The outline of the first embodiment is to discriminate between the case where noise is conspicuous and the case where it is not conspicuous based on the quantization noise generated when the information amount is expanded in the video information amount compression apparatus. In the case of being conspicuous, the size of the DCT block is made smaller than that in the case of being inconspicuous, and the information amount compression is performed to obtain the information amount compression signal. Then, when the information amount expansion is performed, the information amount compressed signal is decoded using the discrimination information for discriminating the size of the DCT block.

(Image information amount compression device) In FIG.
The main part of the image information amount compression apparatus is a main encoding unit AA including first and second encoding units AA1 and AA2 for compressing an information amount of an input image signal, and the first and second encoding units. Control means BB for outputting a control signal for selecting signals output from AA1 and AA2, respectively, and the first and second
Of the encoding means AA1 and AA2 respectively, and the multiplexing means CC for adding ID information for discriminating them. Hereinafter, the main encoding means AA will be described in order.

The DCT block 1a obtained by extracting pixel data of 8 vertical and 8 horizontal from the input video signal is the first and second two-dimensional DC in the first and second encoding means AA1 and AA2.
It is supplied to the T converters 1 and 11, respectively. Here, the DCT block shown in FIG. 2A will be considered as an example of the DCT block 1a, and another numerical example will be described below based on this numerical example. It should be noted that the numbers representing the digital data in FIG. 9A are quantized by 8 bits,
The value 127 can be taken. Further, the input video signal may be any signal representing image information such as a component signal including a Y signal, an RY signal, and a BY signal.

The first two-dimensional D in the first encoding means AA
The CT converter 1 is a two-dimensional D converter for pixel data of 8 vertical and 8 horizontal pixels.
The transform coefficient block 1b is obtained by CT transform and is supplied to the first quantizer 2.

Here, in order to explain the two-dimensional DCT transformation by using mathematical expressions, the two-dimensional D for N × N pieces of data f (j, k) is described.
The CT transform is shown in Equation 3 below, and the inverse transform is shown in Equation 4.

[0026]

[Equation 3]

[0027]

[Equation 4] Then, F (u, v) is called a conversion coefficient, and in particular, u = v = 0 is called a DC conversion coefficient, and other cases are called an AC conversion coefficient. The calculation performed by the first two-dimensional DCT converter 1 described above is a case where N = 8.

Then, the quantizer 2 quantizes the transform coefficient block 1b by quantizing each value of the quantization matrix shown in FIG.
Are supplied to the first encoder 10.

Then, the first encoder 10 encodes the quantized block 2c to obtain a first encoded signal 10a.
To one input of the selector 21 in the multiplexing means CC.

The first encoder 10 may be any one that performs well-known encoding, for example, the one-dimensional predictor 5 and the first and second Huffman codes as described in the prior art. It comprises digitizers 6, 8, a zigzag scanner 7 and a multiplexer 9.

On the other hand, the second 2 in the second encoding means AA
The dimensional DCT converter 11 first divides the DCT block 1a into four first to fourth sub-blocks Sa to Sd, each consisting of four vertical and four horizontal pixel data. This is illustrated in FIG. 2 (A). Then, the first to fourth sub blocks Sa to S
Each of the d's is two-dimensionally DCT-transformed to obtain a second transform coefficient block 11a, and these are supplied to the second quantizer 12.
Note that this two-dimensional DCT transformation is N
= 4.

Then, the second quantizer 12 converts the second transform coefficient block 11a into a quantization matrix composed of four vertical coefficients and four horizontal coefficients, and a scaling factor S = 0.
Quantization is performed by dividing the value multiplied by 5 as the quantization step. Then, the second quantization block 12a thus obtained is supplied to the second encoder 13. This quantization matrix is shown in FIG. 2B, the second sub-block Sb is two-dimensionally DCT-transformed and quantized, and the result is shown in FIG. 2C, and the fourth sub-block Sd is two-dimensionally DCT-transformed. The quantized result is shown in FIG. It should be noted from Expression 3 that each value obtained by performing the two-dimensional DCT conversion on the first and third sub-blocks Sa and Sc becomes “0”.
It is clear that the quantized values of these are also "0".

The second encoder 13 performs entropy coding similarly to the first encoder 10 described above. At this time, the DC conversion coefficients of the four sub-blocks are averaged to obtain the DC conversion coefficient as the DCT block.
The DC conversion coefficients corresponding to the second and fourth sub-blocks Sb and Sd have the values in the upper left corners of FIGS. 7C and 7D, and the DC corresponding to the first and third sub-blocks Sa and Sc. The conversion coefficient becomes "0". Therefore, the DC conversion coefficient of the entire DCT block can be obtained by the following equation 5.

[0034]

[Equation 5] The DC conversion coefficient of the entire DCT block thus obtained is one-dimensionally predicted by the DC conversion coefficient of the entire adjacent DCT block, and the resulting prediction error is Huffman coded.

Next, the second to fourth sub blocks Sb to Sd
For the DC conversion coefficients of, the respective calculation results are replaced with the difference values between these calculation results and the DC conversion coefficients of the entire DCT block. The processing results corresponding to the first to fourth sub-blocks Sa to Sd obtained by performing the above processing are shown in FIGS.

Then, the data "7" in the upper left corner of FIG.
Except for 5 ", zigzag scanning is performed in the matrix in the order of (E) → (F) → (G) → (H) in the figure. Then, conversion coefficients that are not“ 0 ”are grouped, After the run-length coding is performed on the transform coefficient of "0", the Huffman coding is performed.

The second coded signal 13 thus obtained
a is applied to one input of a selector 21 in the multiplexing means CC.

The control means BB will be described below.
The two-dimensional HPF output block 14 obtained by performing the spatial HPF by supplying the DCT block 1a to the three-dimensional HPF 14
The value a is supplied to the first absolute value converter 15.

Here, an example of the coefficient of the two-dimensional HPF 14 is shown in FIG. The coefficient “8” surrounded by a thick frame in FIG. 1I represents the coefficient of the target pixel, and the other coefficient “−1” represents the coefficient of the peripheral pixels of the target pixel. .. Then, each piece of data forming the two-dimensional HPF output block 14a is obtained by multiplying the digital data of the target pixel by 8 and adding 9 pieces of data obtained by multiplying the digital data of the peripheral pixels by -1. I am trying. still,
If the target pixel is at the edge of the DCT block and there are not eight pixels in the periphery, the coefficient of the target pixel is made equal to the number of effective pixels. The two-dimensional HPF output block 14a thus obtained is shown in FIG.

The first absolute value converter 15 takes an absolute value for each data constituting the two-dimensional HPF output block 14a, and takes this absolute value as the first and second threshold values, respectively.
The signals are supplied to the second comparators 16 and 18, respectively.

The first comparator 16 compares the respective data of the absolute-valued two-dimensional HPF output block 11a with the first threshold value set to a value larger than the second threshold value, and compares the first data value with the first threshold value. If it is larger than the threshold value of "1", otherwise "0" is obtained as a logical value for each data, and this is used as the first comparator output signal 16a for the first area determiner 1
Supply to 7. It should be noted that the first threshold is a value that allows detection when there is a high spatial frequency component with a relatively large amplitude, and the second threshold is a value that is determined when there is a high spatial frequency component with a relatively small amplitude. It is a value that can be detected.

The first area discriminator 17 shown in FIG. 3A is 1 when the logical value of each pixel data of the first comparator output signal 16a is "1" in the DCT block.
If there is one, it will detect this.

The set / reset FF 23 shown in FIG.
Since the first comparator output signal 16a is supplied to the set input S of, the output signal of the set / reset FF 23 becomes high level after the first comparator output signal 16a becomes high level. Further, since the reset input is supplied with the DCT block switching signal 23b which remains at the high level for a certain period every time the DCT block is switched, the output signal of the set / reset FF 23 becomes low after the DCT block switching signal 23b becomes the high level. It becomes a level. Then, this output signal is supplied to the D-type FF 24, and this is supplied to the DCT.
It is latched by the block switching signal 23b to obtain the first area discriminator output signal 17a.

In this way, it is possible to detect a high spatial frequency component having a relatively large amplitude in the DCT block.

On the other hand, the second comparator 18 outputs “1” when the absolute value of each data of the two-dimensional HPF output block 14 a is larger than the second threshold value, and “0” in other cases. Is obtained as a logical value for each data, and this is used as the second comparator output signal 18a for the second area determiner 1
Supply to 9.

The second area discriminator 19 shown in FIG. 3B has a second area in the first to fourth areas shown in FIG. 2J.
If there is at least one logical value of the comparator output signal 18a for each pixel data of "1", this is detected. The same components in FIG. 3B as those in FIG. 3A are designated by the same reference numerals.

First, the second comparator output signal 18a
~ The first to fourth area signals 25a to 28a, which are "1" when in the fourth area and "0" otherwise, are supplied to one input of the second to fifth AND circuits 25. ~ 2
To the other input of 8 respectively. Then, the output signals obtained by the logical product here are the second to fifth signals to which the DCT block switching signal 23b is supplied to the reset input.
Are supplied to the set inputs of the set / reset FFs 29 to 32, respectively. The output signals of the second to fifth set / reset FFs are supplied to the 4-input NAND circuit 33, the logical product of these is taken and inverted, and then the DCT block switching signal 2
The second area determiner output signal 19a is obtained via the D type FF 34 latched by 3b.

In this way, if there is at least one flat part in four sub-blocks, this can be detected,
In such a case, the second area discriminator output signal 19a is "1".
In other cases, it becomes "0".

Then, the first AND circuit 20
The logical product of the output signals 17a and 19a of the second area discriminator is calculated, and the logical product is supplied to the selector 21 and the ID adder 22 as the control signal 20a.

To explain the multiplexing means CC, the selector 21 selectively outputs the first and second encoded signals 10a and 13a for each DCT block using the control signal 20a. That is, when the control signal 20a is “0”, 8 × 8 DC
The first coded signal 10a obtained by using T, and the second coded signal 13a obtained by using 4 × 4 DCT in the case of “1”
Is supplied to the ID adder 22 as a selector output signal 21a.

Then, the ID adder 22 controls the control signal 2
An information amount compression output signal 22a obtained by adding discrimination information for discriminating the first encoded signal 10a and the second encoded signal 13a based on 0a is supplied to a transmission line (not shown). (Image Information Decompression Device) To explain the image information amount expansion device with reference to FIG. 4, the information amount compression output signal 22a is supplied to the second selector 40, where it is added by the ID adder 22. The separated discrimination information signal 40a and video information signal 40b are separated and supplied to the ID decoder 41 and the third selector 42, respectively.

The ID decoder 41 decodes the discrimination information signal 40a to obtain the ID decoder output signal 4
1a is supplied to each of fourth and fifth selectors 45 and 55 described later.

On the other hand, the third selector 42 is the DCT described above.
The DC conversion coefficient signal 42b and the AC conversion coefficient signal 42a corresponding to the DC conversion coefficient of the entire block are supplied to the first and second Huffman decoders 44 and 43, respectively.

The first Huffman decoder 44 decodes the DC conversion coefficient signal 42b and outputs the decoded DC conversion coefficient signal 44a to one input of first and second multiplexers 49 and 48 which will be described later. Supply each.

The second Huffman decoder 43 supplies the decoded AC conversion coefficient signal 43a obtained by decoding the AC conversion coefficient signal 42a to the fourth selector 45. The fourth selector 45 corresponds to the first selector 21 in the information amount compression device, and decodes the alternating current in accordance with the discrimination information signal 40a for discriminating between 4 × 4 DCT and 8 × 8 DCT. The conversion coefficient signal 43a is transferred to the first and second inverse zigzag scanners 47, 4
6 is to be supplied.

The first inverse zigzag scanner 47 is for restoring the zigzag scanning for each 1DCT block performed by the first encoder 10 described above, and is a memory for rearranging data. Etc. And
The first inverse zigzag scanner output signal 47a thus restored is supplied to the other input of the first multiplexer 49.

The second inverse zigzag scanner 46 is for returning the zigzag scanning for each sub-block performed by the second encoder 13 to the original, and is a memory for rearranging data. Etc. Then, the second inverse zigzag scanner output signal 46a thus restored is supplied to the other input of the second multiplexer 48.

The first multiplexer 49 multiplexes the decoded DC conversion coefficient signal 44a and the first inverse zigzag scanner output signal 47a to obtain the first 8x8 data. The multiplexer output signal 49a is transferred to the first dequantizer 50.
To the 8 × 8 inverse DCT converter 51. The first inverse quantizer output signal 50a obtained by performing the inverse quantization here is supplied to the 8 × 8 inverse DCT converter 51. Then, the first decoded DC obtained by the inverse DCT conversion
The T block signal 51a is supplied to the fifth selector 55.

On the other hand, the second multiplexer 48 outputs the decoded DC conversion coefficient signal 44a and the second inverse zigzag scanner output signal 4
The second dequantizer 5 outputs the second multiplexer output signal 49a composed of 4 × 4 × 4 data obtained by multiplexing 6a and 6a.
2 and the second inverse quantizer output signal 52a obtained by performing inverse quantization here is supplied to the 4 × 4 inverse DCT converter 53. Then, 4 × 4 inverse DC obtained by inverse DCT conversion
The T converter output signal 53a is supplied to the block converter 54, where a sub block composed of 4 × 4 data is converted into a DCT block composed of 8 × 8 data. Incidentally, it goes without saying that the DC conversion coefficient of each sub-block is subjected to an operation reverse to that at the time of compression. The second decoded DCT block signal 54a thus obtained is supplied to the fifth selector 55.

Then, the fifth selector 55, based on the ID decoder output signal 41a, outputs the first and second decoded DCTs.
The information amount expansion signal 55a obtained by selecting the block signal is output to a video device (not shown).

In this way, the video information amount expansion device outputs the information amount compressed signal 22a output from the video information amount compression device.
The data amount expansion signal 55a can be obtained by subjecting the information amount expansion to.

Now, in order to explain the usefulness of the present embodiment, the case of dividing into sub-blocks and the case of not dividing into sub-blocks will be compared with reference to FIG.

FIG. 5 (A) shows the structure of FIG.
Information amount expansion signal 55a applied to the DCT block of (A)
Is. On the other hand, FIG. 5B shows the result of performing the 8 × 8 DCT transform of the DCT block of FIG. 2A and quantizing the value obtained by multiplying the quantization matrix by the scaling factor S = 0.5 as the quantization step. Also, this is 8 × 8 inverse DC
FIG. 5C shows the result of T-transformation and dequantization. still,
In the 8 × 8 DCT transform and the 8 × 8 inverse DCT transform, N = 8 in the equations 3 and 4 described above.

Comparing FIG. 7A and FIG. 7C, it can be seen that noise due to quantization noise is smaller in FIG.

Further, the number of bits after compression of the transform coefficient matrix for each sub-block shown in FIGS. 2E to 2H, which has been entropy coded by the above-described method, and FIG.
When the transform coefficient matrix of (B) is entropy-coded and the number of bits after compression is compared, the former is 248.
The latter becomes 281 bits. However, the two-dimensional Huffman encoding table shown in FIG. 6 is used, and EOB (end of block) is Run / Si.
It is assumed that ze corresponds to 0/0 and 16 run lengths correspond to F / 0.

Therefore, the amount of information is also smaller when divided into sub-blocks and subjected to 4 × 4 DCT conversion. That is, it is possible to suppress the noise of the pixel data at the time of expansion without increasing the amount of information after compression so much. (Second Embodiment) In the first embodiment, the first and second encoding means AA1 and AA2 are separately provided in the video information amount compression apparatus. However, since both have many common functions, it is possible to reduce the circuit scale by adaptively switching a part of the circuit operation. Therefore, in the second embodiment, the circuit is made common and the circuit scale is reduced. (Video Information Amount Compressing Device) The video information amount compressing device according to the second embodiment will be described with reference to FIG. 7. What is different from FIG. 1 is the main encoding means AA and the multiplexing means CC. Since the control means BB has the same configuration, the same reference numerals are given and the description thereof is omitted.

DC is applied to the delay circuit 56 in the main encoding means AA.
T block 1a is provided, where DCT block 1a
Is delayed, and the timing is adjusted with the control signal 20a for controlling the switching between the 4 × 4 DCT and the 8 × 8 DCT. Then, the delayed DCT block 1a is converted into a two-dimensional DCT by the first and second one-dimensional DCT converters 57 and 58.
The conversion is applied.

First and second one-dimensional DCT converters 57 and 5
The reason why the two-dimensional DCT transform can be performed by the cascade connection of 8 is as follows. In the DCT transform expressed by the above-described Equations 3 and 4, the product-sum operation regarding j and the product-sum operation regarding k are independently Can be done, and u, v
Is also the same. Therefore, the result of performing the product-sum operation on k may be subjected to the product-sum operation on j. That is, 2
The dimensional DCT conversion can be realized by the cascade connection of the one-dimensional DCT conversion. This one-dimensional DCT conversion is given by the following expression 6 according to the expression of expression 3.

[0069]

[Equation 6] Then, an example of a signal flow graph of an algorithm for performing N = 8 and N = 4 in Expression 6 is shown in FIG.
Shown in (A) and (B). In the figure, a confluence point (for example, XX) from the left side represents addition or subtraction (those having a "-" sign). Further, the circle symbol represents a multiplier, "C" in the circle symbol represents COS, the number in the upper row following it represents the numerator in which the coefficient π is omitted, and the number in the lower row represents the denominator. For example, YY is “COS (4π / 1
6) ”represents the operation of multiplying the input by the output.

Comparing FIG. 7A and FIG. 7B, the algorithm of FIG. 7B (when N = 4) is the same as that of FIG. ), It is possible to share the configuration. The algorithm shown in FIG. 7 is an example of a high-speed algorithm, and there are many common parts in other well-known algorithms, and the configuration can be shared.

In this way, the first one-dimensional DCT converter 57 performs the product-sum operation on k, and the second one-dimensional DCT converter 58 performs the product-sum operation on j based on the result. The one-dimensional DCT is extended to the two-dimensional DCT.

Both the first and second one-dimensional DCT converters 57 and 58 can be realized by the block diagram shown in FIG. 9, for example. Then, each component has a control signal 2
Since the switching operation is performed by sharing the 4 × 4 DCT and the 8 × 8 DCT by 0a, the number of control units in the data rearrangement circuit is mainly increased compared to either configuration, and both are configured independently. It can be realized with a simple configuration.

The output signal of the second one-dimensional DCT converter 58 is supplied to the quantizer 59. This quantizer 5
Also for 9, since the quantization matrix for the 4 × 4 DCT transform (FIG. 2B) is a part of the quantization matrix for the 8 × 8 DCT transform (FIG. 14), the configuration can be simplified by sharing it.

Regarding the DC conversion coefficient 59a, the first
As described in detail in the embodiment, it is divided into sub blocks and 4 ×
In the case of 4DCT conversion, the first to fourth sub-blocks Sa
It is necessary to find the average of the DC conversion coefficients of ~ Sd. Further, for the second to fourth sub-blocks Sb to Sd, it is necessary to obtain the difference value between the DC conversion coefficient of each sub-block and the above average, and the difference value is supplied from the DC conversion coefficient calculator 61 to the zigzag scanner 60. The signal that is showing represents this. Since these calculations are binary data, they can be configured only by the adder and the shift register, and the configuration does not become so complicated.

As for the zigzag scanner 60, the 4 × 4 DCT conversion and the 8 × 4 DCT conversion are performed as described in the first embodiment.
Although different scanning is performed by the x8 DCT conversion, the memory circuit, which is a main part of the configuration, can also be used, so that the configuration can be simplified.

Also, the third and fourth Huffman encoders 6
For 3 and 64, the same encoder may be used for the 4 × 4 DCT transform and the 8 × 8 DCT transform, so that the configuration can be simplified. Furthermore, a selector is not required for the multiplexing means CC.

In this way, the control signal 20a causes 4
Since the common configuration is adaptively switched when performing the x4 DCT conversion and the 8 × 8 DCT conversion, it is possible to provide a video information amount compression device having a simple configuration. (・ Video information amount expansion device) In the video information amount expansion device as well as the video information amount compression device, 4 × 4 inverse DCT conversion and 8
Since the common configuration can be adaptively switched when performing the x8 inverse DCT conversion, the video information amount expansion apparatus of the second embodiment will be described with reference to FIG.
In the figure, the same components as those in FIG. 4 are designated by the same reference numerals, and the description thereof will be omitted. That is, the configurations from the input to the first and second Huffman decoders 44 and 43 are the same, and the configurations thereafter are different.

The first Huffman decoder 44 supplies the decoded DC conversion coefficient signal 44a to the one-dimensional predictive decoder 71, where the signal obtained by the one-dimensional predictive decoding is the DC conversion calculator 72 and the third multiplex decoder. And the rectifier 73.

On the other hand, the second Huffman decoder 43 supplies the decoded AC conversion coefficient signal 43a to the data rearrangement memory 70. The data rearrangement memory 70 corresponds to the first and second inverse zigzag scanners 47 and 46 and corresponds to the 4 × 4 inverse DCT.
In the case of conversion or 8 × 8 inverse DCT conversion, the operation of inverse zigzag scanning is switched by using the ID decoder output signal 41a. Then, the signal subjected to the inverse zigzag scanning in the data rearrangement memory 70 is the DC conversion coefficient calculator 72.
Is supplied to the third multiplexer 73.

The DC conversion coefficient calculator 72 obtains the DC conversion coefficient applied to the sub-block, and calculates the DC conversion coefficient.
Supply to 3. The third multiplexer 73 multiplexes the input signal using the ID decoder output signal 41a and supplies the multiplexed signal to the third dequantizer 74.

The third dequantizer 74 operates by switching the dequantization matrix in the case of 4 × 4 inverse DCT transform or 8 × 8 inverse DCT transform using the ID decoder output signal 41a. The signal obtained by performing the inverse quantization is the first
To the one-dimensional inverse DCT converter 75.

The first one-dimensional inverse DCT converter 75 and the second one-dimensional inverse DCT converter 76 are respectively connected to the first and second one-dimensional DCT converters 57 and 58 (shown in FIG. 7). Corresponding, switching operation is performed in the case of 4 × 4 inverse DCT conversion or 8 × 8 inverse DCT conversion using the ID decoder output signal 41a, and two-dimensional DCT conversion is performed by cascade connection of one-dimensional DCT conversion. The final output information amount expansion signal 55a is obtained.

In this way, even in the video information amount expansion device, the common configuration can be shared in the case of 4 × 4 inverse DCT conversion or 8 × 8 inverse DCT conversion, so that the overall configuration can be simplified.

Although the video information amount compression device and the video information amount decompression device have been individually described in the above-mentioned embodiment,
Of course, an information amount compression / decompression device that integrates both may be used.

When one frame of a television video signal is compression-encoded as one image, interfield motion is suppressed in order to suppress interference caused by interfield motion based on the interlaced structure of the video signal. When 8 "is detected and" no motion "is detected, 8x8 DCT is selected, and when" motion "is detected, 4 lines are collected for each field and 4x8 DCT x 2 is selected. A method for suppressing motion obstruction is known. Even in such a case, the present embodiment can be applied, and in the case of "no motion", 4
Of course, the divided 4 × 4 DCT may be selected, and in the case of “with motion”, the divided 4 × 4 DCT may be selected.

[0086]

As described above, according to the present invention, the vertical M
DCT conversion of the input pixel block consisting of N pixels and N pixels horizontally,
The first coded signal obtained by quantizing and coding and P vertical pieces,
A second coded signal obtained by DCT-transforming, quantizing, and coding each of the horizontal Q sub-blocks is selectively multiplexed based on the high spatial frequency component of the input pixel block, Since the information amount compressed signal can be obtained by adding the discrimination information for discriminating the first encoded signal and the second encoded signal, the information amount is expanded without increasing the information amount so much. There is an effect that it is possible to provide a video information amount compression or information amount compression / expansion device that outputs an information amount compression signal in which noise based on quantization noise is suppressed.

Further, when the information amount compression signal is subjected to the information amount expansion, the information amount compression signal having the first and second encoded signals can be decoded based on the discrimination information. There is an effect that it is possible to provide a video information amount compression or information amount compression / decompression device which outputs an information amount decompression signal in which noise based on noise is suppressed.

[Brief description of drawings]

FIG. 1 is a block diagram illustrating a main part of a video information amount compression device according to a first embodiment.

FIG. 2 is a diagram showing a numerical example for explaining a main part of the video information amount compression device of the first embodiment.

FIG. 3 is a circuit diagram for explaining an example of first and second area determiners.

FIG. 4 is a block diagram for explaining a main part of the video information amount expansion device of the first embodiment.

FIG. 5 is a diagram showing numerical examples for comparing the first embodiment with a conventional technique.

FIG. 6 is a table showing an example of Huffman coding.

FIG. 7 is a block diagram for explaining a main part of a video information amount compression device according to a second embodiment.

FIG. 8 is a signal flow graph for explaining a high-speed algorithm of one-dimensional DCT conversion.

FIG. 9 is a block diagram for explaining first and second one-dimensional DCT converters.

FIG. 10 is a block diagram for explaining a main part of a video information amount expansion device according to a second embodiment.

FIG. 11 is a block diagram of a conventional still image compression apparatus.

FIG. 12 is a diagram for explaining the operation of the zigzag scanner.

FIG. 13 is a diagram for explaining image quality deterioration due to compression / expansion using DCT due to quantization noise.

FIG. 14 is a diagram for explaining a quantization matrix.

[Explanation of symbols]

1a DCT block (input pixel block) 10a First coded signal 13a Second coded signal 20a Control signal 22a Information amount compression signal 27 Two-dimensional LPF 40a Discrimination information signal (discrimination information) AA Main encoding means AA1 First Coding means AA2 second coding means BB control means CC multiplexing means Sa to Sd first to fourth sub-blocks (sub-blocks)

Claims (2)

[Claims] 1. A first encoding means for outputting a first encoded signal obtained by DCT transforming, quantizing and encoding an input pixel block obtained by dividing a video signal, and the input pixel block. A plurality of sub-blocks obtained by dividing
Main encoding means having a second encoding means for outputting a second encoded signal obtained by transforming, quantizing, and encoding, and detecting and obtaining a high spatial frequency component of the input pixel block. Control means for outputting the control signal for selecting the first coded signal and the second coded signal for each of the input pixel blocks based on the signal, and the control means based on the control signal. Multiplexing for outputting an information amount compressed signal obtained by adding discrimination information based on the control signal to a selection signal obtained by selecting one of the first encoded signal and the second encoded signal. A video information amount compression or information amount compression / decompression device comprising: 2. A video information amount decompression or information amount compression / decompression device for decompressing the information amount compressed signal according to claim 1, wherein the first amount is based on the decoded discrimination information. A video information amount decompression or information amount compression / decompression device for decoding the information amount compression signal having a coded signal and the second coded signal.