KR100197373B1 - Object coding apparatus - Google Patents

Abstract

The present invention relates to an object-based encoding apparatus for compensating motion for each object by dividing a video by moving objects and estimating a motion expression parameter for each moving object. The image segmentation unit 21 for dividing each image, the block segmentation unit 22 for dividing the image divided for each moving object by a predetermined block unit, and detecting a block including an image fragment among the blocks divided in the predetermined block unit. An image fragment block detector 23 for determining a number of filter taps for filtering each one-way line of a predetermined block according to the number of pixels forming an image fragment in a block including the image fragment; A low pass filtering of each one-way line according to the number of taps determined by the first tap number determination unit 24 to expand an area in the one direction 1 FIR expansion filter unit 25, a second tap number determination unit 26 for determining the number of filter taps for filtering the line in the other direction perpendicular to the one direction with respect to the predetermined block extended in the one direction, the second And a FIR extension filter unit 27 for extending the region in the other direction by low-pass filtering each other line of the predetermined block according to the number of taps determined by the first tap number determination unit 26. .

Description

Object coding device

1 is a diagram illustrating a conventional block-by-block encoding technique.

2 is a detailed block diagram showing an embodiment of the present invention.

3 is a view for explaining a filtering process according to the present invention.

* Explanation of symbols for main parts of the drawings

21: video divider 22: block divider

23: image fragment block detection unit 24, 26: tap number determination unit

25,27: FIR Expansion Filter

The present invention is an object-specific encoding apparatus for compensating motion for each object by dividing a video by moving object and estimating a motion expression parameter for each moving object, and in particular, an MF (Model) that does not sufficiently express a signal change from the motion parameter. Failure) relates to an object-specific encoding device for encoding the pixel information of the region.

Recently, studies on low-rate video encoding techniques that can be accommodated ahead of the standardization of the Motion Picture Experts Group (MPEG) 4 have been actively conducted. In order to encode low transmission of continuous images including moving objects, a data reduction technique using moving information is used. This is essential.

Meanwhile, conventional video encoding techniques are based on local block motion regardless of the shape and global motion of the moving object.

Therefore, at low data rates, by applying block-by-block motion compensation coding, image quality deterioration such as blocking, corner blur, and spot phenomenon appears in the playback image.

In MPEG4, unlike the conventional method which considers only the statistical characteristics of video signals as a new coding scheme to solve the above problems, it focuses on subjective image quality based on human visual characteristics and drastically reduces the amount of transmission data. A video encoding technique is proposed, which is one of the new coding techniques, which divides a video by moving objects and estimates motion expression parameters for each moving object. to be.

Object-specific encoding technique first divides the input image into moving objects of independent motion and extracts contour information, motion information, and color information by estimating the motion information of the divided object, and then extracts the contour information, motion information, Color information is selectively encoded and transmitted for each moving object.

In this object-specific coding scheme, most regions can be motion compensated by motion coefficients, so only motion information and contour information need to be transmitted.

However, an area that does not sufficiently express the signal change from the motion coefficients is generated. Since the MF area cannot compensate for the movement, it is necessary to transmit not only the contour information but also the brightness information.

That is, when the encoding is performed by the object-specific encoding method, pieces of arbitrary shapes are generated and not necessary to encode the N * M rectangular image.

Therefore, conventionally, there are GOT (General Orthogonal Transform) proposed by Gilge and iterative coding scheme presented by Kaup and Aach as a method for encoding MF region brightness information. Although it shows good performance because of the encoding, these techniques require a lot of time and computation to find the basis function because the orthogonal basis functions vary depending on the contour shape.

In addition, a block-by-block transform encoding method for encoding MF region brightness information is performed by filling zero, an average value, and a mirror of the region in a region outside the desired object region, and then converting it. If described with reference to:

First, the zero stuffing technique is the simplest way of filling zero into an area 2 other than the object area 1 as shown in FIG. do. Because of this, there is a problem that the compression rate is much lower when encoding because a lot of unwanted high frequency components are included.

As methods proposed to solve the problems of the zero filling technique described above, an average value filling method (FIG. 1 (b)) and a mirror image expansion method which increase coding efficiency by filling other components other than zero outside the MF region (FIG. 1C).

The average value filling method is a method of filling an average value of brightness values in the object area 3 into an area 4 other than the object area, and the mirror image expansion method is used to extend the brightness value of the object area 5 outside the object area ( 6), it has the advantage of reducing the high frequency component because it prevents the abrupt change of the signal to the boundary area, but the coding efficiency is inferior because the encoding is performed after filling the unwanted information only in the object area. There was a problem.

SUMMARY OF THE INVENTION The present invention has been made to solve the above problems, and an object thereof is to provide an encoding apparatus for an object that can increase encoding efficiency by performing region expansion by filtering an image fragment region having an arbitrary shape in horizontal and vertical directions. .

According to an aspect of the present invention, there is provided an image divider for dividing an image of a frame unit input by moving objects of independent motion, a block divider for dividing an image divided by each moving object by a predetermined block unit, and An image fragment block detection unit for detecting a block including an image fragment among each block divided by a set block unit, and a filter tap number for filtering each one-way line of the preset block according to the number of pixels forming an image fragment in the block including the image fragment A first FIR extension filter part configured to low-pass each of the one-way lines and expand the area in the one direction according to the number of taps determined by the first tap number determiner, and to expand the area in the one direction Determining the number of filter taps for filtering the line in the other direction perpendicular to the one direction for the predetermined block And an FIR extension filter unit for low-pass filtering each other line of the predetermined block according to the number of taps determined by the first tap number determination unit, and extending the region in the other direction. .

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

FIG. 2 is a detailed configuration diagram for implementing the present invention. An image splitter 21 for dividing an input frame unit image into moving objects having independent motions and an image divided by moving objects in units of 8 * 8 blocks. The image fragments detected by the block dividing unit 22 for dividing, the image fragment block detecting unit 23 for detecting blocks including the image fragments among the blocks divided by 8 * 8 blocks, and the image fragment block detecting unit 23 The number of taps determined by the number of taps determiner 24 and the number of taps determiner 24 for determining the number of filter taps for filtering each horizontal line of an 8 * 8 block according to the number of pixels forming an image fragment in the containing block. The FIR (Finite Impulse Response) expansion filter unit 25 and the FIR expansion filter unit 25 expand the area in the horizontal direction by low-pass filtering each horizontal line of the 8 * 8 block including the image fragment according to Zone extended 8 * 8 blocks Each vertical number of 8 * 8 blocks extended in the horizontal direction according to the number of taps determined by the number of taps determining unit 26 and the number of taps determining unit 26 for determining the number of filter taps for filtering each vertical line The FIR extension filter portion 27 extends the region in the vertical direction by low pass filtering the line.

The operation description of the embodiment of the present invention configured as described above is as follows.

First, the image dividing unit 21 divides an input image of a frame unit for each moving object of independent motion, estimates motion information of the divided object, extracts contour information and color information, and then extracts the contour information and color information to the block dividing unit 22. The block dividing unit 22 divides the image divided by the moving object into 8 * 8 block units.

Therefore, the image fragment block detector 23 detects a block including an image fragment among blocks divided in units of 8 * 8 blocks, and provides the block to the tap number determiner 24, and the tap number determiner 25 performs image fragments. The number of filter taps for filtering each horizontal line of 8 * 8 blocks is determined and provided to the FIR extension filter unit 26 according to the number of pixels constituting the image fragment in the block including the FIR extension filter unit 26. ) Filters the horizontal lines of each 8 * 8 block including the image fragments according to the determined number of taps to expand the area in the horizontal direction.

For example, if each pixel value of one horizontal line of 8 * 8 blocks is 0,0, a1, a2, a3,0,0,0 as shown in FIG. 3 (a), the number of filter taps at this time is right It will have 4 taps and 3 taps to the left. On the other hand, the FIR extension filter unit 26 is composed of eight FIR filters according to the number of taps.

That is, each FIR filter has M filter taps to the left and N filter taps to the right, N + M is 8, and the taps of each FIR filter are (N = 1, M = 7), (N = 2, M = 6 ), (N = 3, M = 5), (N = 4, M = 4), (N = 5, M = 3), (N = 6, M = 2), (N = 7, M = 1 )

Therefore, the tap number determination unit 25 determines four taps to the right and three taps to the left and provides the taps to the FIR extension filter unit 26, and the FIR extension filter unit 26 determines the number of taps determined by the tap number determination unit 25. Low-pass filtering with a tap number, i.e., (N = 4, M = 3), results in a1, a1 + a2, a1 + a2 + a3, a1 + a2 + a3, a1 + a2 as shown in FIG. + a3, a1 + a2 + a3, a2 + a3, a3.

The tap number determination unit 27 determines the number of filter taps for filtering each vertical line for an 8 * 8 block in which each horizontal line is extended in the horizontal direction by the FIR extension filter 26 as shown in FIG. 3 (b). The vertical tap number determination is the same as the horizontal tap number determination method described above. In addition, according to the number of taps in the vertical direction determined by the tap number determining unit 27, the FIR expansion filter unit 28 low-pass filters each vertical line of 8 * 8 blocks that are extended in the horizontal direction to expand the area in the vertical direction. The FIR extension filter unit 28 is also composed of eight different tabs, like the FIR extension filter unit 26, and performs the same function as the above-described FIR extension filter unit 26, but is characterized in that it performs in the vertical direction. You can do it.

As described above, the preferred embodiment of the present invention filters the image fragments in the horizontal direction and again filters them in the vertical direction, but it should not be understood that the object-specific encoding scheme of the present invention is necessarily limited thereto. Even if the first filtering in the vertical direction rather than the direction, and then the horizontal filtering again, those skilled in the art will know that the result will be the same.

As described above, the present invention has an effect of improving encoding efficiency by encoding an object by regularly filtering image fragments in a horizontal and vertical direction.

Claims (2)

An image divider 21 for dividing an input frame-based image by moving objects of independent motion, a block divider 22 for dividing an image divided by each moving object by a predetermined block unit, and the preset block unit An image fragment block detection unit 23 for detecting a block including an image fragment among each block divided into?, And a filter tap number for filtering each one-way line of the preset block according to the number of pixels constituting the image fragment in the block including the image fragment A first FIR extension filter unit 25 which expands an area in the one direction by low pass filtering each of the one-way lines according to the number of taps determined by the first tap number determination unit 24 to determine a number of taps determined by the first tap number determination unit 24. ), The second tap number determination unit 26 for determining the number of filter taps for filtering the other direction lines perpendicular to the one direction with respect to the predetermined block extended in the one direction. A first number of taps the object-specific coding device with a FIR expansion filter portion 27 to expand the area in the other direction by low-pass filtering each of the other direction line of the predetermined block in accordance with the determined number of taps by the determination section 26. The object-specific encoding device of claim 1, wherein the predetermined block is an 8 * 8 block.