JPH0669812A - Information compression-encoding device and information expansion-decoding device - Google Patents

Abstract

PURPOSE:To enable high-efficiency encoding without complicating circuit configuration by performing a prescribed processing to information signals and transmitting a first code word and a second code word. CONSTITUTION:Input image information aa is supplied to a block dividing means 1 and the output signal is supplied to a DCT transformer 2 to perform two-dimensional DCT at every block. As the result of two-dimensional DCT, a transformation coefficient F(U, V) is derived. Further, a quantizing means 3 quantizes the transformation coefficient according to a prescribed method. This quantized transformation coefficient is encoded by using a '0' pattern detecting means 4 and a variable length encoding means. At the means 4, the nonsignificant coefficient of a requantized transformation coefficient corresponding to a previously prepared pattern is encoded. At the means 5, remaining transformation coefficients excluding the non-significant coefficient previously stored in the means 4 from requantized arrangement are encoded.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an information compression coding apparatus and an information decompression decoding apparatus for performing high-efficiency coding and decoding of voice and images.

[0002]

2. Description of the Related Art With the digitization of images and sounds, high efficiency coding technology has become important. Orthogonal transform coding is an effective means of high efficiency coding. As the orthogonal transform, for example, DCT (discrete cosine
Transformation) is used. A two-dimensional DCT is often used for image information.
(DCT, two-dimensional DCT, edited by Hiroshi Harashima, edited by The Television Society, "Image Information Compression", Ohmsha (199
1), p. See 104. Hereinafter, it is referred to as Document 1. ) An example of a coding method using the two-dimensional DCT will be described below. First, the image is divided into blocks of 4 pixels × 4 pixels, two-dimensional DCT is performed on each block, and the resulting transformation coefficient F (u, v); u, v = 0, 1, 2, 3 Get a pair. It is assumed that this set of conversion coefficients is virtually arranged as shown in FIG. As shown in the figure, from F (0,0) to F
(3, 3) are arranged in order. F (0,0) is called a DC conversion coefficient, and the other coefficients are called AC conversion coefficients. As shown by the arrows in the figure, the AC conversion coefficients are arranged so as to represent higher spatial frequency components toward the lower right. Next, the transform coefficients thus obtained are requantized, for example, in the coarse quantization step for the high frequency component transform coefficient and in the fine quantization step for the low frequency component transform coefficient. The requantized transform coefficients are sequentially encoded using a variable length code such as Huffman code.

As a coding method after requantization, in addition to the method of simply Huffman coding all transform coefficients as described above, as a method capable of further efficient coding, reference 1, p. There is a method described in 200. The characteristic parts of the method are summarized below. First, the AC conversion coefficients of the set of conversion coefficients virtually arranged as shown in FIG. 6 are zigzag scanned as shown in FIG. 7, and the coefficients obtained by requantization in that order are encoded. Encoding is performed as follows. First, the number of coefficients having a value of 0 (hereinafter referred to as non-significant coefficients) between a coefficient having a non-zero value (hereinafter referred to as a significant coefficient) and the already-signed significant coefficient immediately before that is calculated, Is the run length. When the run length is 0, the value of the significant coefficient is Huffman coded. When the run length is 1 or more, a code indicating that the run length code continues next and a code obtained by Huffman coding the run length are output, and then the value of the significant coefficient is Huffman coded. Further, when the encoding of the last significant coefficient at the time of scanning is completed as shown in FIG. 7, a code indicating the end of the block is output.

Alternatively, p. 290 to JPEG
As described as the (Joint Photographic Experts Group) method, there is also a method of performing two-dimensional Huffman coding by combining the run length of the non-significant coefficient and the value of the significant coefficient.

[0005]

Among the above methods, in the first-mentioned method of simply performing variable-length coding on the values of transform coefficients, it is necessary to perform variable-length coding on all values 0 even for insignificant coefficients. Yes, the coding efficiency is relatively poor. Also, the method of using the run length of the non-significant coefficient described below does not completely remove the redundancy of the position and value of the transform coefficient, and a simple method that can further improve the coding efficiency is desired. It was

[0006]

The present invention provides the following configurations in order to solve the above problems.

First obtained by subjecting information signal to orthogonal transformation
Transforming means for outputting the transforming coefficient, the quantizing means for outputting the second transforming coefficient obtained by subjecting the first transforming coefficient to predetermined quantization, and the first transforming coefficient for the second transforming coefficient. First encoding means for generating a first code word that substantially matches the pattern of the non-significant coefficient and that corresponds to a specific pattern of the second non-significant coefficient including the pattern of the first non-significant coefficient; The first
And a second coding means for coding the second transform coefficient that has not been coded by the coding means to generate a second codeword.

An information decompression / decoding apparatus for the information compression / encoding apparatus according to claim 1, comprising decoding means for decoding based on the first codeword and the second codeword. And an information decompression / decoding device.

[0009]

1 is a block diagram of an embodiment of an encoding / decoding apparatus according to the present invention, FIG. 2 is a conceptual diagram for explaining the output of a quantizing means, and FIGS. 3 and 4 are "0" patterns. FIG. 5 is a conceptual diagram for explaining the position pattern of the non-significant coefficients provided in advance by the detecting means, and FIG. 5 is a conceptual diagram for explaining the encoding processing of the variable length encoding means. Embodiments will be described below with reference to the drawings.

In FIG. 1, the coding / decoding apparatus is composed of the configurations 1 to 5 on the code side, the transmission line 6, and the configurations 7 to 11 on the decoding side. As will be described from the code side, an input image information signal aa which is a digital signal is supplied to a block dividing means 1 which divides into a block of 4 pixels × 4 pixels, and an output signal (information signal) thereof is a two-dimensional DCT for each block. DC to apply
It is supplied to the T converter 2 (conversion means).

Then, the transform coefficient F as a result of the two-dimensional DCT
(U, v); u, v = 0, 1, 2, 3 (first transform coefficient) is obtained. It is assumed that this set of conversion coefficients is virtually arranged as shown in FIG. Further, it is assumed that the transform coefficient is requantized by the quantizing means 3 by a predetermined method, and as a result, the array shown in FIG. 2A (or FIG. 2B) is obtained. This requantized transform coefficient (second transform coefficient) is used by the "0" pattern detecting means 4 (first coding means) and the variable length coding means 5 (second coding means). It is encoded as follows.

First, the "0" pattern detecting means 4 extracts the position pattern of the non-significant coefficient from the arrangement shown in FIG. 7A (or FIG. 7B) and then the position pattern shown in FIG. )) Pattern (first non-significant coefficient pattern). Incidentally, the hatched portion in FIG. 7C (or FIG. 7D) shows the position of the non-significant coefficient.

On the other hand, a non-significant coefficient position pattern (second non-significant coefficient pattern) is set in advance as shown in FIG.
The position patterns of (D) and FIGS. 4A to 4D are set to “0”.
The pattern detecting means 4 is prepared, and these position patterns are typical ones that have a high probability of occurrence through statistical processing. The hatched portions in FIGS. 3 and 4 are the positions of non-significant coefficients. The "0" pattern detecting means 4 assigns 000 to 111 as shown in FIGS. 3 and 4 to each pattern as a 3-bit code word.

The position pattern of the non-significant coefficient shown in FIG. 2C (or FIG. 2D) is compared with the position patterns of FIGS. 3 and 4, and the position pattern of FIG. ) (Or (D) in the same figure), the non-significant coefficient positions (hatched parts in FIGS. 3 and 4) are included in the parts not included in the position pattern of the non-significant coefficients (parts not hatched in the drawing). 2) (or FIG. 2D), the number of nonsignificant coefficients that is the most or close to the portion (hatched portion in the figure) included in the position pattern of the nonsignificant coefficient in FIG. The "0" pattern detecting means 4 selectively outputs the one having the position (hatched portion in FIGS. 3 and 4).

The hardware that performs this selective output processing is a ROM that uses the position of the non-significant coefficient in FIG. 2C (or FIG. 2D) as an address input and outputs the position pattern in FIGS. This can be achieved by preparing (read only memory). Alternatively, in the position patterns of FIGS. 3 and 4, the top row is the 0th from the right (assuming that there is no non-significant coefficient position) or the first position is the non-significant coefficient position. The third row is the 0th, first or second position from the right, the third row is the first, second, third or fourth position from the right, the fourth row is Since the second, third, or fourth positions from the right are insignificant coefficient positions, if the 0th to 4th positions in each row are represented by 1 bit, 2 bits, 2 bits, and 2 bits, respectively, Since the ROM address can be represented by a total of 7 bits, the hardware can be simplified.

Here, as compared with FIG. 2C, FIG.
(Or the pattern of FIG. 4D with respect to FIG. 2D) is selected. Therefore, the codeword 100 (or 11
1) is output. In this way, the "0" pattern detecting means 4 encodes the non-significant coefficient of the requantized transform coefficient, which corresponds to the pattern prepared in advance.

Next, the variable length coding means 5 is used in FIG.
From the arrangement of (A) (or FIG. 2 (B)) to FIG. 4 (A)
(Or, the non-significant coefficients at the hatched positions in FIG. 4D are excluded, and the remaining conversion coefficients (the coefficients at the unhatched positions in FIG. 5A (or FIG. 5B)) are coded. To convert. The encoding is as shown in FIG.
The remaining transform coefficients are read in the following order from the left side of the top row of (B)).

A, C, 0, D, B, 0, 0, 0, E, F ---
(1) (or A, C, B, D ---
(2)) Then, the values of the read transform coefficients are encoded in order, including the value 0 of the insignificant coefficient, by variable length coding such as Huffman coding. Alternatively, the non-significant coefficient may be encoded using the run length as in the conventional example.

As a result of the above, the codes to be transmitted are the 3-bit codeword 100 (or 111) and the codeword in which the numerical sequence of (1) (or (2)) is variable-length coded.
By transmitting the 3-bit codeword at a timing before the variable-length coded code, both the coding side and the decoding side can know the number of the variable-length coded codes (for example, the 3-bit code When the number of words is 100, the number of variable-length coded codes is 10, which corresponds to the number of unhatched portions in FIG. 4A.) Therefore, it is not always necessary to transmit the code indicating the end of the block. Absent.

The coded signal 5a thus obtained is supplied to the "0" pattern decoding means 7 on the decoding side via the transmission line 6. This transmission line 6 is a magnetic tape recording / reproducing system when the encoding / decoding device is a digital VTR, and an optical recording / reproducing system when it is a magneto-optical recording / reproducing device. It should be noted that the encoding / decoding device may be divided into an image information compression encoding device and an image information decompression decoding device, and in this case, it may be a recording / reproducing only device. In addition, the transmission line 6
May be radio waves.

On the decoding side, the "0" pattern decoding means 7
The 3-bit code word 100 (or 111) is decoded to prepare the position pattern of the virtual insignificant coefficient of FIG. 4 (A) (or FIG. 4 (D)).

Next, in the variable length decoding means 8, (1)
(Or (2)) variable-length coded codeword is decoded to obtain the numerical sequence of (1) (or (2)), and each numerical value is converted to the numerical value shown in FIG. 4 (D)) Insert it in the unhatched position. As a result, the signal obtained by reproducing the arrangement of FIG. 2A (or FIG. 2B) is supplied to the inverse quantizer 9.

Then, a block of image data consisting of 4 pixels × 4 pixels is obtained through the inverse quantizing means 9 and the inverse DCT converter 10, which are in a complementary relationship with the quantizing means 3 and the DCT converter 2, respectively. The output image information signal bb obtained by re-combining this with the screen synthesizing means 11 is output to a transmission line (not shown).

Although a 3-bit fixed-length code word is assigned to the patterns shown in FIGS. 3 and 4 in the above embodiment, it goes without saying that a variable-length code may be used as the code word.

Of course, it is not always necessary to use the patterns of FIGS. 3 and 4 as typical position patterns of non-significant coefficients, and for example, a method of using a plurality of patterns in descending order of appearance probability can be considered. .

Further, in this embodiment, the pattern of FIG. 4 (A) is selected as the pattern having the most non-significant coefficient positions in the portion included in the position pattern of the non-significant coefficients of FIG. 2 (C). In order to simplify the wear, FIG. 4C may be selected in which the number of non-significant coefficients is the smallest but is close to that. It is also possible to adopt a configuration in which only the DC conversion coefficient is encoded differently as in the conventional example.

[0027]

As described above, according to the configuration of the present invention,
Of the transform coefficients obtained after requantizing the transform coefficients obtained as a result of the orthogonal transform, most of the non-significant transform coefficients are prepared in advance out of the second position pattern which is the position pattern of the representative non-significant coefficients. One of the position pattern of the representative non-significant coefficient and the code word obtained by encoding the other transform coefficient excluding the non-significant coefficient included in the pattern by a predetermined method. Since it is configured to transmit a code word representing one, there is an effect that it is possible to provide an information compression encoding device that can perform encoding with higher efficiency and that does not require complicated hardware for encoding.

Further, since there is a decoding means for decoding based on the first code word and the second code word, there is an effect that it is possible to provide an information decompression decoding device corresponding to the above information compression coding device.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of an encoding / decoding device according to the present invention.

FIG. 2 is a conceptual diagram for explaining an output of a quantizing unit.

FIG. 3 is a conceptual diagram for explaining a position pattern of non-significant coefficients provided in advance by a “0” pattern detection means.

FIG. 4 is a conceptual diagram for explaining a position pattern of non-significant coefficients provided in advance by a “0” pattern detection means.

FIG. 5 is a conceptual diagram for explaining an encoding process of a variable length encoding means.

FIG. 6 is a conceptual diagram for explaining a two-dimensional DCT.

FIG. 7 is a conceptual diagram for explaining zigzag scanning.

[Explanation of symbols]

 2 DCT converter (converting means) 3 quantizing means 4 "0" pattern detecting means (first coding means) 5 variable length coding means (second coding means) 7 "0" pattern decoding means ( Decoding means) 8 Variable length decoding means (decoding means)

Claims (2)

[Claims] 1. A first obtained by subjecting an information signal to orthogonal transformation.
Transforming means for outputting the transforming coefficient, a quantizing means for outputting a second transforming coefficient obtained by subjecting the first transforming coefficient to predetermined quantization, and a first transforming coefficient for the second transforming coefficient. First encoding means for generating a first code word that substantially matches the pattern of the non-significant coefficient and that corresponds to a specific pattern of the second non-significant coefficient including the pattern of the first non-significant coefficient; An information compression encoding device, comprising: second encoding means for encoding the second transform coefficient that has not been encoded by the first encoding means to generate a second codeword. . 2. An information decompression / decoding apparatus for the information compression / encoding apparatus according to claim 1, comprising decoding means for decoding based on said first codeword and said second codeword. An information decompression / decoding device characterized by: