KR0180173B1 - An object coder - Google Patents

Abstract

The present invention relates to an object encoder, comprising: an image splitter (10) for blocking an outline of an object into 8x8 pixel blocks; A DCT unit 20 for converting the 8x8 pixel block into an 8x8 frequency coefficient block; A zigzag scanning unit (30) for zigzag scanning the 8x8 frequency coefficient block and performing run length encoding on the 8x8 frequency coefficient block; A vertical scanning unit 40 for vertically scanning the run length coding of the 8x8 frequency coefficient block and outputting the run length encoding; A horizontal scanning unit (50) for horizontally scanning the run length coding of the 8x8 frequency coefficient block and outputting the same; An edge detection unit (60) for detecting an edge of the object blotted with the 8x8 pixel block and outputting a scanning selection signal according to the direction of the edge; According to the scanning selection signal is configured by solving the selection unit 70 for selecting and outputting the output of the zigzag scanning unit 30, the vertical scanning unit 40 and the horizontal scanning unit 50, Since DCT is performed on the contour portion of the blocked object, scanning is performed by selecting an appropriate scanning direction according to the edge direction of the object, thereby reducing the amount of image data.

Description

An object coder

1 is a view showing a state in which the contour of an object is blocked.

2 is a schematic diagram of an object encoder according to the present invention.

3 is a schematic diagram illustrating a scanning method according to the present invention.

* Explanation of symbols for main parts of the drawings

10: video splitter 20: DCT unit

30: zigzag scanning unit 40: vertical scanning unit

50: horizontal scanning unit 60: edge detector

70: selector 80: variable length coder (VLC)

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an object encoder. In particular, the frequency coefficient produced by blocking the contour portion of an object and performing a discrete cosine transform (DCT) is scanned in various directions according to the edge direction of the object, thereby reducing the amount of image data. An object encoder for reducing.

At present, the multimedia era, which combines and integrates multiple media such as computer, communication, and broadcasting, has been developed in various ways. Core technologies that support such multimedia include digitalization and digital image compression.

To briefly discuss the necessity of such digital image compression technology, the current NTSC television (TV) signal has a data rate of 180 Mbits per second. If you put it in a CD (Compact Disk, capacity: about 6.25Gbit), it will be about 35 seconds, so you will not be able to store enough video data with this information storage capacity. Therefore, a technique for storing a lot of image data on a single CD has been developed, which is a digital image compression technique.

In summary, the digital image compression reduces the amount of data required to display an image by removing spatial and temporal redundancy of the image.

Meanwhile, the MPEG-2 video compression method is based on two basic techniques.

First, we use motion estimation and compensation on a block-by-block basis to reduce temporal redundancy. Second, we use DCT compression to reduce spatial redundancy.

That is, there are three types of picture types defined in MPEG-2, 1, P, and B-pictures. Among them, I (Intra coded) pictures are simply DCT-coded without using motion compensation. In the P (Predicted coded) picture, motion compensation is performed based on I or another P-picture and DCT-coded the remaining difference. In addition, B (Bidirectionally predicted coded) pictures use motion compensation like P-pictures, but unlike P-pictures, they compensate for motion from two frames before and after on the time axis.

This motion compensation is based on 16 × 16 blocks.

In addition, the DCT coding is an orthogonal transform having a fast algorithm, which has a performance close to optimum for many kinds of images, and the DCT basic function facilitates the effective use of the visual characteristic criteria. It is being used to reduce space redundancy.

This DCT is based on 8 8 blocks.

However, in the conventional coding method that processes blocks or macroblocks using mathematical statistical characteristics of an image as described above, the lower the bit rate, that is, the higher the compression rate, the more severely the block shape and the blurring of the boundary are generated. Visually annoying.

Accordingly, in view of the fact that human vision is sensitive to boundaries, an object coding method for dividing and encoding an image based on an object has been developed.

This conventional object coding method codes a contour of an object of irregular shape and gray levels in the object are separately fixed by a method such as DCT.

In this case, as shown in FIG. 1, it is necessary to fill a blank portion other than an object with a value before performing the DCT. Usually, the DCT is performed after filling the blank portion with '0'.

However, if the blank portion is filled with '0' and DCT is performed as described above, high frequency coefficients are generated, so that a large number of bits are required to code the frequency coefficients.

Accordingly, the present invention has been made to solve the above problems, by reducing the amount of image data by scanning the frequency coefficient produced by blocking the contour portion of the object and then performing a DCT in different directions according to the edge direction of the object. It is an object of the present invention to provide an object encoder.

In order to achieve the above object, the object encoder according to the present invention comprises: an image splitter which blocks an outline portion of an object into an 8 × 8 pixel block; A DCT unit for converting the 8x8 pixel block into an 8x8 frequency coefficient block; A zigzag scanning unit which scans the 8 × 8 frequency coefficient block by zig zag and executes run length encoding of the 8 × 8 frequency coefficient block; A vertical scanning unit which performs run length coding by vertically scanning the 8x8 frequency coefficient flocks and outputting the same; A horizontal scanning unit which horizontally scans the 8x8 frequency coefficient block by run length encoding and outputs the same; The output of the zigzag scanning unit is detected when the edge of the object is diagonal and the output of the vertical scanning unit is selected when the edge direction is horizontal, and the output of the horizontal scanning unit is selected when the edge direction is vertical. An edge detector for outputting a scanning selection signal; And a selection unit for selecting and outputting one of the zigzag scanning unit, the vertical scanning unit, and the horizontal scanning unit according to the scanning selection signal.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

2 is a schematic diagram of an object encoder according to the present invention, wherein the object encoder comprises: an image splitter 10 for blocking an outline portion of an object into 8 × 8 pixel blocks; A DCT unit 20 for converting the 8x8 pixel block into an 8x8 frequency coefficient block; A zigzag scanning unit (30) for zigzag scanning the 8x8 frequency coefficient block and performing run length encoding on the 8x8 frequency coefficient block; A vertical scanning unit 40 for vertically scanning the run length coding of the 8x8 frequency coefficient block and outputting the run length encoding; A horizontal scanning unit (50) for horizontally scanning the run length coding of the 8x8 frequency coefficient block and outputting the same; An edge detector (60) which detects an edge of the object blocked by the 8x8 pixel block and outputs a scanning selection signal in accordance with the detected edge direction; A selector (70) for selecting and outputting one of the zigzag scanning unit (30), the vertical scanning unit (40), and the horizontal scanning unit (50) according to the scanning selection signal; For the length coded codeword output from the selector 70, short bits per code are allocated to codewords having a high generation rate, and long bits per code are assigned to codewords having a low occurrence probability. By removing the statistical redundancy so that the average length of the code is close to the entropy, and including a variable length encoder (VLC, 80) for transmitting to the receiving side (not shown).

The edge detector 60 detects an edge of an object and outputs the zigzag scanning unit 30 when the direction of the edge is oblique, and outputs the vertical scanning unit 40 when the direction of the edge is horizontal. In the case of the vertical, the scanning selection signal for outputting the horizontal scanning unit 50 is selected to the selector 70.

If described in more detail the operation and effect of the object encoder according to the present invention configured as described above.

First, when the image splitter 10 blocks the outline portion of the image into 8 × 8 pixel blocks and outputs them to the DCT unit 20, the DCT unit 20 outputs the received 8 × 8 pixel blocks by 8 × 8 frequency coefficients. The blocks are converted into blocks and output to the zigzag scanning unit 30, the vertical scanning unit 40, and the horizontal scanning unit 50, respectively.

The zigzag scanning unit 30, the vertical scanning unit 40, and the horizontal scanning unit 50 each run-length-code the received 8 × 8 frequency coefficient block to the selector 70.

The selector 70 is a run length coded codeword output from one of the zigzag scanning unit 30, the vertical scanning unit 40, and the horizontal scanning unit 50 according to the scanning selection signal provided by the detector 60 yesterday. These are output to the variable length encoder (80).

That is, as shown in FIG. 3, the selector 70 selects and outputs a run length coded codeword that is zigzag scanned and output after performing DCT when the edge of the object is oblique, and the edge of the object is a vertical line. In this case, the run length coded codeword which is horizontally scanned and output after the DCT is selected and output. When the edge of the object is a horizontal line, the run length coded codeword which is vertically scanned and output after the DCT is selected and output.

For example, the DCT frequency coefficient is shown in Table 1 below,

Table 1

First, if the contour edge of an 8 × 8 blocked object is oblique, run by zigzag scanning in the order (276, 137, 59, 89, -94, 51, -12, 25, -35, 39 ...) Select and output the length coded codeword, and if the contour edge of the second 8 × 8 block-wound object is a vertical line (276, 59, 89, 39, 7, -13, -12, -7, 137 ...) Scan run-length coded codewords by horizontal scanning in order, and thirdly, when the contour edge of 8 × 8 blocked object is a horizontal line (276, 137, 51, -12, -8) , 2, 1, -1, -6, 59 ...) and then run-length coded codewords are selected and output.

At this time, if the blank part other than the object is filled with '0' and DCT is performed, a high frequency coefficient is generated, and a large number of bits are required to code the frequency coefficient, but as described above, the scanning direction is determined according to the edge direction of the object. By varying, the amount of video data can be reduced.

On the other hand, the length-encoded codewords selectively scanned as described above are assigned short bits per code to codewords having a high probability of occurrence in the variable length encoder 70 and codes for codewords having a low probability of occurrence. The long bits are allocated to remove the statistical redundancy so that the average length of the code is close to entropy and then transmitted to the receiving side (not shown).

As described above, the object encoder according to the present invention has the effect of reducing the amount of image data since DCT of the contour portion of the blocked object and then scanning by selecting a suitable scanning direction according to the edge direction of the object. .

Claims (1)

An image splitter 10 for blocking an outline of an object into 8 × 8 pixel blocks; A DCT unit 20 for converting the 8x8 pixel block into an 8x8 frequency coefficient block; A zigzag scanning unit 30 which scans the 8 × 8 frequency coefficient block by zig zag and executes run length encoding of the 8 × 8 frequency coefficient block; A vertical scanning unit 40 for vertically scanning the run length coding of the 8x8 frequency coefficient block and outputting the run length encoding; A horizontal scanning unit (50) for horizontally scanning the run length coding of the 8x8 frequency coefficient block and outputting the same; When the edge of the object is detected and the direction of the edge is diagonal, the output of the zigzag scanning unit 30 is output. When the direction of the edge is horizontal, the output of the vertical scanning unit 40 is vertical. An edge detection unit 60 for outputting a scanning selection signal for allowing the output of the horizontal scanning unit 50 to be selected; And a selection unit 70 for selecting and outputting one of the zigzag scanning unit 30, the vertical scanning unit 40, and the horizontal scanning unit 50 according to the scanning selection signal. Object encoder.