JPH04360469A - Adaptive scan coding circuit - Google Patents

Abstract

PURPOSE:To attain most effective compression coding in response to a picture when a coding coefficient subjected to orthogonal transformation is coded in the coding and decoding employing both the orthogonal transformation and the entropy coding. CONSTITUTION:A discrete cosine transformation circuit 2 applies discrete cosine transformation to picture information in response to the frequency component and 1st, 2nd, 3rd scan circuits 3, 4, 5 with different scan sequence are used to rearrange the result into plural linear picture information sets and 1st, 2nd, 3rd coding circuits 6, 7, 8 connecting respectively to them encode the result. The coded picture information is given to 1st, 2nd, 3rd data length counter circuits 10, 11, 12, from which code quantity is counted and the least value is detected by a comparator circuit 13 from the result of count and the result of detection is outputted. Then the picture information in response to the result of detection is selected by a selection circuit 9 and selected picture information and scan information are outputted mutiplexedly from a multiplexer circuit 14.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to highly efficient encoding of images, and relates to a composite encoding circuit that uses orthogonal transformation and entropy encoding together.

0002

[Prior Art] Discrete cosine transform (DC), which is one of orthogonal transforms, is used to encode and compress image information with high efficiency.
There is a composite encoding that uses a combination of Huffman encoding, which is a type of entropy encoding (for example, "Television Society Journal" published by the Television Society of Japan, 1989).
Vol. 43, No. 10, ``Still Image Coding Method,'' pages 1145 to 1155, and ``Image Coding Algorithm II,'' published by the Television Society of Japan, 1990.
Vol. 44, No. (See article 2, “Still Image Coding Method,” pages 153 to 161).

[0003] The overall information compression rate when using this composite coding of discrete cosine transform and Huffman coding depends on how efficiently the coding coefficients after the discrete cosine transform are coded. In other words, discrete cosine transformation simply represents image information, which is two-dimensional information, with multiple cosine functions with different frequency components in the vertical and horizontal directions, and rearranges the levels of each frequency component in frequency order and encodes it. It is only output as a coefficient, and there is no information compression itself. Therefore, the overall information compression rate is
It is determined by how the coding coefficients for each frequency component obtained by discrete cosine transform are coded.

[0004] Generally, the compression rate of the entire encoding is improved at the stage of linear quantization and dimensional transformation before encoding for the encoding coefficients output from the discrete cosine transform. The procedure will be explained below.

[0005] Linear quantization is quantization by dividing the coding coefficient using a table (quantization coefficient table) in which predetermined numerical values are arranged, and the desired compression is achieved by setting the values of the quantization coefficient table. You can get the rate. In addition, the coding coefficients output from the discrete cosine transform are arranged in the order of their frequency components (low frequency components, which indicate monotonous changes in the image, are to the left or above, and high frequency components, which indicate fine changes in the image, are to the right or below). Therefore, the compression ratio can be further improved by utilizing this characteristic to remove high frequency components that are difficult to recognize due to the visual characteristics of a person placed in the lower right direction.

Image information compressed by linear quantization is rearranged into one-dimensional coding coefficients for encoding. For rearrangement from two-dimensional to one-dimensional, zigzag scanning, in which two-dimensional coding coefficients are scanned in a zigzag pattern from the upper left to the lower right, as shown in FIG. 2, is generally used. This zigzag scan scans the coding coefficients expressed in two dimensions sequentially from the upper left direction to the lower right direction along the line connecting the upper right and lower left as shown in Figure 2. This is to rearrange the encoding coefficients. By scanning in a zigzag pattern, when 0s are concentrated in the lower right direction of the two-dimensional encoded coefficients, for example, when compressing high frequency components by linear quantization,
Replace the number followed by (0 run length) with a specific sign,
The amount of information can be further compressed.

[0007] The encoding coefficients rearranged from two dimensions to one dimension by zigzag scanning are encoded by Huffman encoding. In Huffman encoding, one-dimensional coding coefficients are replaced with codes consisting of shorter bit strings in descending order of frequency of occurrence, and coefficient values with lower frequency of occurrence are replaced with codes consisting of longer bit strings. In other words, Huffman encoding is encoding that compresses the amount of information by indicating the information that occurs more frequently with a shorter code.

[0008]

[Problems to be Solved by the Invention] When compressing and encoding image information using the above-mentioned composite encoding, the compression efficiency is limited when the image to be compressed and encoded is a monotonous image, or when high-frequency components are compressed. 2 is an acceptable image, etc.
The same bit data (
A high compression ratio can be obtained when there are many 0 or 1). That is, the compression efficiency increases as the run length generated by zigzag scan for two-dimensional encoded coefficients increases.

[0009] However, images with few consecutive 0s in the direction of the straight line connecting the upper right and lower left in the two-dimensional coding coefficients,
For example, images that cannot be compressed into high-frequency components during linear quantization because deterioration in image quality is not allowed and high-frequency components remain in the coding coefficients, or discrete images such as those shown in Figure 4 (1) and Figure 5 (1) Coding coefficients during cosine transformation (see Fig. 4 for (1) in Fig. 4)
(2), Fig. 5 (2)) with respect to Fig. 5 (1) shows that in images where the occurrences are concentrated in the horizontal or vertical direction, a large zigzag scan in the straight line connecting the upper right and lower left The occurrence of run lengths is small, and the effect of compressing information cannot be obtained.

The present invention has been made in view of the above-mentioned problems, and it is an object of the present invention to provide an adaptive scan encoding circuit that can perform compression encoding most effectively depending on the image.

[0011]

[Means for Solving the Problems] The present invention provides orthogonal transformation means for orthogonally transforming image information according to frequency components, and an apparatus for scanning image information transformed by the orthogonal transformation means in a plurality of different extraction orders. A scanning means for outputting a plurality of dimensional image information together with scan information in the order in which it is taken out, and encoding one-dimensional image information that has the least amount of information when encoded among the plurality of one-dimensional image information output from the scanning means. and an image encoding means for outputting the scan information together with the scan information.

0012

[Operation] The orthogonal transformation means orthogonally transforms the image information according to the frequency components, and the scanning means scans the image information transformed by the orthogonal transformation means in a plurality of different extraction orders to convert it into one-dimensional image information. The image encoding means encodes the one-dimensional image information that has the least amount of information when encoded among the plurality of one-dimensional image information outputted from the scanning means. The scan information is output together with the scan information.

[0013]

[Embodiment] As an adaptive scan encoding circuit according to an embodiment of the present invention, FIGS. 1 to 5 are illustrated using discrete cosine transformation, which is one of orthogonal transformations, and Huffman encoding, which is one of entropy encoding, as examples. I will explain while referring to it.

FIG. 1 is a schematic diagram of an embodiment of an adaptive scan encoding circuit according to the present invention. In FIG. 1, 1 is a block division circuit that divides input image information into blocks of a predetermined size, and 2 is a block division circuit that performs discrete cosine transformation on the image information of each predetermined size divided by the block division circuit 1. A discrete cosine transform circuit 3 serves as an orthogonal transform means that performs linear quantization and outputs two-dimensional encoded coefficients. 3 is a discrete cosine transform circuit that performs linear quantization and outputs the two-dimensional encoded coefficients output from the discrete cosine transform circuit 2 in the order shown in FIG. A first scan circuit 4 is a scanning means that scans and rearranges the two-dimensional coding coefficients in one dimension using a zigzag scan (zigzag scan). A second scan circuit 5 serves as a scanning means for scanning and rearranging into one dimension, and 5 scans the two-dimensional coding coefficients output from the discrete cosine transform circuit 2 in the order shown in FIG. A third scan circuit serves as a scanning means for sorting, and 6 is scanned by the first scan circuit 3 and 1
A first encoding circuit serves as an encoding means for Huffman encoding the encoding coefficients rearranged into dimensions, and 7 is a second encoding circuit.
A second encoding circuit 8 serves as an encoding means for Huffman encoding and outputting the encoded coefficients scanned by the scan circuit 4 and rearranged in one dimension; 10 is a first data length counter circuit that counts the information length of the image information encoded by the first encoding circuit 6; 11 is the second
A second data length counter circuit that counts the information length of image information encoded by the encoding circuit 7; 12 is a third encoding circuit 8;
A third data length counter circuit 13 counts the information length of the encoded image information, and 13 indicates the three types of encoded information amounts ( The first coded under different scanning orders,
A comparison circuit that compares the outputs of the second and third encoding circuits 6, 7, and 8 and outputs scan type information, etc. of the one with the smallest amount of information as a comparison detection signal; 9 is a comparison detection signal from the comparison circuit 13; The first, second and third encoding circuits 6 based on the signal
, 7 and 8, a selection circuit as a selection means for selecting encoded image information having the least amount of encoded information from the encoded image outputs of , 7 and 8;
Reference numeral 4 denotes a multiplexing circuit 14 that multiplexes the scan type and the like in the comparison detection signal from the comparison circuit 13 onto the encoded image information selected by the selection circuit 9.

The case where image information is encoded in such a circuit will be explained.

In FIG. 1, image information is input to a block dividing circuit 1 and divided into blocks of a predetermined size. The image information divided into blocks is subjected to discrete cosine transformation for each predetermined block by a discrete cosine transformation circuit 2. The image information subjected to the discrete cosine transform is linearly quantized according to a predetermined quantization coefficient table and output as two-dimensional encoded coefficients. In this linear quantization, a predetermined compression rate can be obtained by setting the quantization table. For example, if you want to give priority to compression rate over image quality, increase the overall setting value of the quantization coefficient table and
A high compression ratio can be obtained by setting a larger setting value corresponding to a high frequency component. At this time, 2 after linear quantization
In the dimension encoding coefficient, most of the encoded high frequency components located in the lower right direction are output as 0. On the other hand, if the image quality is given priority over the compression rate, the deterioration of the image quality can be suppressed by setting the overall setting values of the quantization coefficient table to be small (close to 1). At this time, 2 after linear quantization
Depending on the state of the image to be compressed and encoded, the dimension coding coefficients are applied to the upper left for a monotonous image, to the entire image for a complex image, and to the upper right for an image with many changes in the horizontal direction. , if the image has many changes in the vertical direction, large coefficient values will occur in the lower left direction.

The linearly quantized two-dimensional encoded coefficients are scanned by first, second, and third scan circuits 3, 4, and 5 in the respective scanning orders to become one-dimensional encoded coefficients. At this time, the run length is detected in the scanned coding coefficients and replaced with a predetermined code. 1st, 2nd
, the coding coefficients rearranged one-dimensionally by the third scan circuits 3, 4, and 5 are sent to the first, second, and third scan circuits 3, 4,
, 5, respectively, and the output thereof is encoded by the first, second, and third data length counter circuits 10, 11, and 12 connected to the respective data length counter circuits 10, 11, and 12. The amount of information for each is counted.

The first, second, and third encoding circuits 6, 7, and 8 encode the encoding coefficients that have been made one-dimensional by the first, second, and third scanning circuits 3, 4, and 5. , the more frequently occurring coefficient values are replaced with codes consisting of shorter bit strings, and the less frequently occurring coefficient values are replaced with codes consisting of longer bit strings.

First, second and third data length counter circuits 1
The encoded image information amounts counted as 0, 11, and 12 are compared in a comparison circuit 13, the one with the smallest image information amount is detected, and a comparison detection signal containing the scan information is output. The selection circuit 9 selects the encoded image information corresponding to the comparison detection signal from the comparison circuit 13 as first, second,
It is selected from the third encoding circuits 6, 7, and 8 and output to the multiplexing circuit 14. The multiplexing circuit 14 multiplexes the encoded image information from the selection circuit 9 and the scan information included in the comparison detection signal from the comparison circuit 13 and outputs the multiplexed signal.

In the above embodiment, three different types of first,
Second and third scan circuits 3, 4, 5 and their corresponding 3
Detect the minimum encoded information amount from the encoded information amounts of the first, second, and third encoding circuits 6, 7, and 8, and the respective encoded information amounts of the first, second, and third encoding circuits 6, 7, and 8. Although data length counters 10, 11, and 12 and a comparison circuit 13 are provided to select the encoded information with the least amount of encoded information by encoding in three types of scan orders, this configuration is limited. It's not a thing. For example, the configuration may be such that the maximum value of the run lengths generated in three different types of first, second, and third scan circuits 3, 4, and 5 is detected, and the scanning order in which the maximum value is detected is encoded. It is also possible to count the total number of 0s generated in three different types of first, second, and third scan circuits 3, 4, and 5, and to encode the scan order in which the maximum value is detected. You can also do it. FIG. 3 shows a configuration for detecting the maximum value of run length in each scan order. Note that the same parts as those in FIG. 3 are given the same reference numerals, and the description thereof will be omitted.

15, 16, and 17 are 0-run counters that detect and output the maximum value of consecutive 0s in each of the encoded information made one-dimensional by the first, second, and third scan circuits 3, 4, and 5. 1 to 3 circuits, 18 is a comparison circuit that compares and detects the maximum number of consecutive 0s from the 0 run counters 1 to 3 circuits 15, 16, and 17 and outputs it as a comparison detection signal;
Reference numeral 9 indicates the first, second and third scan circuits 3, selected by the selection circuit 9 based on the comparison detection signal from the comparison circuit 18;
This is an encoding circuit 19 as an encoding means for encoding and outputting the one-dimensional encoding coefficients from 4 and 5.

In the circuit configuration shown in FIG. 3, the amount of code at the time of encoding can be reduced by selecting the coding coefficient that has the most consecutively detected 0s among the one-dimensional coding coefficients obtained by scanning. The scan order that is estimated to be the minimum can be selected. According to this configuration, the circuit configuration can be simplified by using only one encoding circuit.

Furthermore, in the above embodiment, three types of scanning orders were set as scanning means for rearranging the encoded coefficients after linear quantization from two-dimensional to one-dimensional, and the amount of information was compared. The present invention is not limited to this, and the amount of encoded information may be compared in at least two types of scan orders.

Furthermore, in the above embodiments, the orthogonal transform in composite encoding was explained using one of them, the discrete cosine transform, but it is not limited to this, and for example, the Fourier transform, Hadamard transform, Karunen transform, etc. - Orthogonal transformation such as Roubaix transformation may be used. Similarly, entropy encoding has been explained using Huffman encoding as an example; however, it is not limited to this; examples include arithmetic encoding, modified Huffman encoding (MH), and modified read encoding (MR). ), modified
Entropy encoding such as modified read encoding (MMR) may also be used.

[0025]

[Effects of the Invention] As is clear from the above description, the present invention has the following advantages:
When sorting the coding coefficients from the discrete cosine transform from two dimensions to one dimension, the coding coefficients are scanned in multiple ways, and the compression efficiency is determined from among the scanned one-dimensional coding coefficients. Since the encoded information obtained by the superior scan is selected and output, efficient encoding can always be performed regardless of the image condition or the quantization coefficient table settings.

[Brief explanation of drawings]

FIG. 1 is a schematic configuration diagram of an adaptive scan encoding circuit according to an embodiment of the present invention.

FIG. 2 is an explanatory diagram of a conventional zigzag scan.

FIG. 3 is a schematic configuration diagram of an adaptive scan encoding circuit according to an embodiment of the present invention.

FIG. 4 is an explanatory diagram of an example of a scanning state for an image.

FIG. 5 is an explanatory diagram of another example of a scan state for an image.

[Explanation of symbols]

1 Block division circuit 2 Discrete cosine transform circuit 3 First scan circuit 4 Second scan circuit 5 Third scan circuit 6 First encoding circuit 7 Second encoding circuit 8 Third encoding circuit 9 Selection circuit 10 First data length counter circuit 11 Second data length counter circuit 12 Third data length counter circuit 13 Comparison circuit 14 Multiplex circuit

Claims (3)

[Claims] 1. Orthogonal transformation means for orthogonally transforming image information according to frequency components, and scanning the image information transformed by the orthogonal transformation means in a plurality of different extraction orders to obtain one-dimensional image information according to the extraction order. a scanning means for outputting a plurality of pieces of information together with the scan information; and one-dimensional image information having the least amount of information when encoded among the plurality of one-dimensional image information outputted from the scanning means, encoded together with the scan information. An adaptive scan encoding circuit comprising an image encoding means for outputting an image. 2. The image encoding means includes a plurality of encoding means each encoding a plurality of one-dimensional image information output from the scanning means, and each image information encoded by the encoding means. 2. The adaptive scan encoding circuit according to claim 1, further comprising a selection means for selecting image information having the smallest amount of information from among them and outputting the selected image information together with the scan information. 3. The image encoding means includes a selection means for selecting image information with a minimum amount of information when encoding a plurality of one-dimensional image information output from the scanning means; 2. The adaptive scan encoding circuit according to claim 1, further comprising: encoding means for encoding the image information and outputting the encoded image information together with the scan information.