KR100229534B1 - Method and apparatus for coding contour information of image - Google Patents

Abstract

The present invention relates to a method and apparatus for encoding a contour of an image signal for encoding the contour of an object included in the image signal using masking and binary quantization techniques.

To this end, in the present invention, a polygonal approximation of a contour image with a plurality of line segments connecting adjacent vertex points in a plurality of vertex points, and providing N sample points at equal intervals on each line segment, Discrete sine transform the approximation error representing the distance between each sample point, output a masking control signal using the length L of each line segment and the number N of sample lines By masking a predetermined high frequency component to 0 in the approximation error, the quantization and encoding are performed, thereby reducing the amount of data to be coded and transmitted and effectively encoding the contour image.

Description

Method and apparatus for contour coding of video signal {METHOD AND APPARATUS FOR CODING CONTOUR INFORMATION OF IMAGE}

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a video signal encoding system, and more particularly, to a method and apparatus for encoding an image signal, which encodes the contour of an object included in the image signal using masking and binary quantization techniques.

In digital television systems, such as video telephony, teleconference, and high definition television systems, the video frame signal includes a sequence of digital data called pixel values, so that a significant amount of digital data is defined to define each video frame signal. I need data.

However, because the available frequency bandwidth on a typical transmission channel is limited, especially in low bit-rate video signal encoders such as video telephony and teleconferencing systems, a significant amount of data can be obtained using various data compressor methods. Should be reduced.

One encoding method for encoding a video signal of a low transmission encoding system is a so-called object-oriented analysis-synthesis coding technique, and according to the object-oriented analysis and synthesis encoding technique, an input image signal Are divided into objects, and the factors defining the motion, contour and pixel data of each object are processed through different coding channels.

In particular, in processing the contour image of the object, the contour information plays an important role in interpreting and composing the shape of the object. A typical coding method used to represent the contour information is a chain coding method. The above-described chain coding method has a disadvantage in that a significant amount of bits are required to encode the contour, although there is no loss of the contour information.

Accordingly, contour approximation methods such as polygonal approximation and B-spline approximation have been proposed. However, if the above methods are applied, the amount of bits consumed in the contour encoding process can be greatly reduced. However, in the case of the polygon approximation method, the approximated contour image is not accurate. In the case of the B-spline approximation method, Although the contour can be represented more accurately than the polygon approximation method, a high order polynomial is required to reduce the approximation error, which complicates the overall calculation in the image encoder.

Therefore, to solve the above problems, a contour coding method using polygon approximation and discrete sine transform (DST) has been devised.

In the contour encoding method based on polygon approximation and DST, first, a plurality of vertex points located on an outline are determined by using a polygon approximation method, and a line segment connecting each adjacent vertex point is used as a line segment. Approximate the outline. In more detail, as shown in FIGS. 1A to 1D, a plurality of vertex points are first determined on the applied contour image 10. At this time, when the contour image 10 is an open-loop, both end points, A and B are selected as starting vertex points, and when the contour image is a closed-loop, the farthest from the contour is farthest. Two points are selected as the starting vertex point. Thereafter, the two selected vertex points are connected by a line segment, and if the maximum distance D _MAX between the corresponding pixels on the contour 10 from this line segment is greater than the predetermined threshold TH1, that pixel is one vertex point. Is selected. This process is repeated until the maximum distance D _MAX is less than the predetermined threshold TH1. As a result, vertex points A through G are determined on the contour 10.

The number of vertex points for a given contour in the approximation process depends on the predetermined threshold TH1. In other words, if the predetermined threshold TH1 is set lower, more vertex points are selected on the outline to express the outline in more detail, while the amount of data to be transmitted increases and the coding efficiency is lowered. Conversely, if the predetermined threshold TH1 is set high, fewer vertex points are selected on the outline, and the outline is roughened, but the amount of data to be transmitted is reduced and the coding efficiency is improved.

Subsequently, as illustrated by way of example in FIGS. 2A and 2B, N sample points (eg, 8) are selected for each line segment forming an outline and each of the N sample points located on each line segment. The approximation errors are computed sequentially, with (where N represents a positive integer). That is, the distance d1 to d8 or d1 between the sample points S1 to S8 or S1 'to S8' on the line segment connecting the vertex points A and D or C and F in a straight line and the corresponding pixel on the contour segment indicating the contour between the vertex points. 'To d8' is detected as an approximation error. The N sample points are arranged at equal intervals on each line segment as shown. Here, all vertex points on the contour are located on the contour segment, so the approximation error for them is zero. The approximation error for the sample point on each line segment calculated above is converted to the DST coefficients through one-dimensional DST, then encoded according to a conventional encoding method, and then transmitted through a transmission path.

However, as shown in Figs. 2A and 2B, in the polygon approximation process, the distance L1 or L2 between vertex points which are both end points of the line segment is shortened so that the distance is smaller than the number of sample points (for example, 8 points). In this case, a problem occurs that unnecessary information is transmitted.

In addition, since most of the error components generated after the polygon approximation process have a shape similar to that of the first and second kernels of the DST coefficients, it is inefficient to apply the same quantization step size to all the DST coefficients. .

The present invention has been made in view of the above, and an object of the present invention is to filter out some of the DST coefficients of the DST approximation error through masking as necessary, and to perform different quantization steps for the low frequency component and the high frequency component among the DST coefficients. By quantizing in size, it is possible to provide an outline encoding method for reducing the amount of data to be transmitted and also effectively encoding the outline image.

Another object of the present invention is to reduce the amount of data to be transmitted by filtering some of the DST coefficients of the DST approximated error through masking as necessary and quantizing the low frequency component and the high frequency component with different quantization step sizes, respectively. An object of the present invention is to provide a contour encoding apparatus for efficiently encoding a contour image.

According to an aspect of the present invention, there is provided a method of encoding a contour image of an object in a video signal, the method comprising: determining a plurality of vertex points located on the contour image; Polygonal approximation of a contour image with a plurality of line segments connecting adjacent vertex points; Providing N sample points at equal intervals on each line segment; Obtaining an approximation error representing the distance between the contour image and each sample point for N sample points on each line segment; Discrete sine transforming an approximation error for each of the plurality of line segments; A masking control signal is provided by using the distance L of each line segment and the number N of sample points, which are obtained by calculating a distance between two vertex points, which are both end points of each line segment, which are discrete sine transformed. process; Masking a predetermined high frequency component with zero among the discrete sine transformed approximation errors to correspond to the masking control signal; Quantizing the masked output; And encoding the quantized output.

According to another aspect of the present invention, there is provided an apparatus for encoding a contour image of an object in a video signal, the apparatus comprising: means for outputting vertex information by detecting a plurality of vertex points located on the contour image; Means for polygonally approximating the contour image as a plurality of line segments connecting adjacent vertex points using the vertex information; Means for providing N sample points at equal intervals on each line segment connecting the adjacent vertex points; Means for detecting an approximation error indicative of a distance between the contour image and each sample point; Means for discrete sine transforming an approximation error for each detected line segment; Means for outputting a masking control signal using a distance L of each line segment and the number N of sample points obtained by obtaining a distance between two vertex points which are both endpoints of each of the line segments that are discrete sine transformed and integerized; Means for masking a predetermined high frequency component with zero among the discrete sine transformed approximation errors to correspond to the masking control signal; Means for quantizing the masked output; And means for encoding the quantized output.

1A to 1D are diagrams illustrating a process of polygon approximating an outline,

2A to 2B are exemplary views illustrating sampling a polygonal approximated outline;

3 is a block diagram of a contour encoding apparatus for a video signal according to the present invention;

4 is a detailed block diagram of the masking control block shown in FIG.

5A and 5B illustrate a process of calculating and masking a distance between vertex points according to the present invention.

<Description of the code | symbol about the principal part of drawing>

10: outline 100: polygon approximation block

200: sampling & error detection block

300: DST 400: Masking Block

600: masking control block 610: length calculation block

620: Masking area determination block 710: First quantization block

720: second quantization block 800: coefficient encoder

900 vertex encoder 950 channel encoder

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

3 is a block diagram illustrating a configuration of an apparatus for encoding an outline of a video signal according to an exemplary embodiment of the present invention.

When the contour image is applied, the polygon approximation block 100 performs the same approximation process as described with reference to FIGS. 1A to 1D, and then outputs vertex information indicating each vertex point on the contour.

In the sampling & error detection block 200, N sample points are displayed on a line segment connecting each adjacent vertex point using vertex information output from the polygon approximation block 100. As shown in FIG. 2B, an equalization error between the contour segment and the line segment defined by each vertex point is detected using the applied contour image and the sample point. In other words, the distance from the sample point on the line segment to the pixel meeting the contour segment by drawing a line vertically is obtained as an approximation error.

The DST block 300 converts the approximation error detected by the sampling & error detection block 200 into DST coefficients in units of line segments. The DST coefficient is composed of N coefficients having different frequencies, and the N value is determined by the characteristics of the sampling & error detection block 200.

During the process, the masking control block 600 provides a signal for controlling the masking block 400 through the line L20 using the vertex information provided from the polygonal approximation block 100 through the line L10.

4 illustrates a masking control block 600 including a length calculation block 610 and a masking area determination block 620. When the vertex information detected by the polygon approximation block 100 is provided to the length calculation block 610 through the line L10, the length calculation block 610 may measure the distance L between two adjacent vertex points, that is, the length L of the line segment. It obtains as shown to 5a. That is, the distance between the two vertex points V1 and V2 is calculated and, for example, integerized using a rounding method to determine the length L and provide it to the masking area determination block 620. In the masking area determination block 620, a masking control signal for determining the masking area based on the line segment length L obtained by the length calculation block 610 is provided to the masking block 400 through the line L20.

In the above description, the number of valid kernels that can occur when the length L obtained using the distance between two vertex points V1 and V2 shown in FIG. 5A is smaller than N, the number of sample points, becomes (L-1). Since the position of each pixel has an integer value, in the case of L = 3, a frequency component larger than the second kernel after the DST process cannot have a great meaning. Therefore, as shown in Fig. 5B, it is advantageous to perform quantization after the length L is L≤8 and after masking (8-L + 1) high frequency components with zero. Accordingly, the masking area determination block 620 calculates (8-L + 1) and provides it to the masking block 500 through the line L20 as a masking control signal.

On the other hand, when the approximation error applied from the sampling & error detection block 200 to the DST block 300 includes an approximation error of the vertex point corresponding to the end point of each line segment, the masking control block 600 controls the masking control. In the process of calculating the signal, we use (N + 2), which includes two vertex points instead of N, in the calculation.

Referring again to FIG. 3, the masking block 400 outputs as many high frequency DST coefficients as the number corresponding to the masking control signal from the masking control block 600 among the DST coefficients for each line segment provided from the DST block 300. Masking is provided to the first stage of the first quantization block 710 and the second quantization block 720.

That is, when N is 8 and the distance L is 2, the masking control signal is 7 when the DST approximation error does not include the approximation error for the two vertex points constituting the line segment. After masking seven high frequency components hatched out of the eight DST coefficients, the remaining one DST coefficient is quantized and transmitted as an active ingredient. Similarly, when L is 5, the masking control signal becomes 4, so that the masked four high frequency components are masked and only the remaining 4 DST coefficients are quantized as an active ingredient.

The DST coefficients masked through the masking block 400 are quantized through the first and second quantization blocks 710 and 720, respectively, according to an embodiment of the present invention, which is the contour of the contour among the low and high frequency components of the DST coefficients. Since the characteristics are almost determined by the low frequency component, it is more effective to dualize the quantization process in a direction of reducing the error of the low frequency component compared to the high frequency component.

Therefore, among the DST coefficients for the line segment, the quantization step size is determined by Equation 1 for the low frequency component and Equation 2 for the high frequency component.

As described above, TH1 represents a threshold of the distance between corresponding pixels on the contour line from the line segment during the polygon approximation process, and M represents the number of quantization steps, respectively, where M is a positive integer.

According to the results of Equation 1 and Equation 2, since the quantization step size of the low frequency component of the DST coefficient corresponds to half of the high frequency component, the error in the quantization process can be reduced accordingly.

Among the quantization step sizes defined above, the low frequency component corresponds to the first quantization block 710 and the high frequency component corresponds to the second quantization block 720. Therefore, the DST coefficients of the low frequency component and the DST coefficients of the high frequency component output from the masking block 500 are provided to the coefficient encoder 800 after being quantized through the first and second quantization blocks 710 and 720, respectively. The encoder 800 is encoded by a well-known variable length coding (VLC) method and then transmitted to the channel encoder 950.

On the other hand, the vertex information provided by the polygon approximation block 100 is applied to the sampling & error detection block 200 and is applied to the vertex encoder 900 and encoded, and then provided to the channel encoder 950. In operation 950, the output of the vertex encoder 900 is encoded along with the output of the coefficient encoder 800 and then transmitted through the transmission path.

As described above, according to the present invention, by calculating the length of the line segment connecting two adjacent vertex points after the polygon approximation and masking unnecessary DST coefficients, while reducing the amount of data to be transmitted through the transmission line By quantizing the low frequency component and the high frequency component in the DST coefficients using different quantization step sizes, the effect of contour coding can be increased.

Claims (8)

In the method for encoding a contour image of an object in a video signal, Determining a plurality of vertex points located on the contour image; Polygonal approximation of a contour image with a plurality of line segments connecting adjacent vertex points; Providing N sample points at equal intervals on each line segment; Obtaining an approximation error representing the distance between the contour image and each sample point for N sample points on each line segment; Discrete sine transforming an approximation error for each of the plurality of line segments; A masking control signal is provided by using the distance L of each line segment and the number N of sample points, which are obtained by calculating a distance between two vertex points, which are both end points of each line segment, which are discrete sine transformed. process; Masking a predetermined high frequency component with zero among the discrete sine transformed approximation errors to correspond to the masking control signal; Quantizing the masked output; Encoding the quantized output Contour encoding method of a video signal comprising a. The contour coding method of claim 1, wherein the masking control signal is defined as N − (L−1), and is regarded as “0” when the result is less than zero. The method of claim 1, wherein the quantization process of the masked output, And quantizing the low frequency components in the masked output corresponding to each of the line segments, and the process of quantizing the high frequency components in the masked output. 4. The method of claim 3, wherein the quantization step size for quantizing the high frequency component is twice the step size for quantizing the low frequency component. An apparatus for encoding a contour image of an object in a video signal, Means for detecting a plurality of vertex points located on the contour image and outputting vertex information; Means for polygonally approximating the contour image as a plurality of line segments connecting adjacent vertex points using the vertex information; Means for providing N sample points at equal intervals on each line segment connecting the adjacent vertex points; Means for detecting an approximation error indicative of a distance between the contour image and each sample point; Means for discrete sine transforming an approximation error for each detected line segment; Means for outputting a masking control signal using a distance L of each line segment and the number N of sample points obtained by obtaining a distance between two vertex points which are both endpoints of each of the line segments that are discrete sine transformed and integerized; Means for masking a predetermined high frequency component with zero among the discrete sine transformed approximation errors to correspond to the masking control signal; Means for quantizing the masked output; Means for encoding the quantized output Contour encoding apparatus of a video signal comprising a. The contour encoding apparatus of claim 5, wherein the masking control signal is defined as N − (L−1), and is regarded as “0” when the result is less than zero. The method of claim 5, wherein the quantization means, Means for quantizing the low frequency components of the masked output corresponding to each of the line segments and means for quantizing the high frequency components of the masked output. 8. The contour encoding apparatus of claim 7, wherein the quantization step size for quantizing the high frequency component is twice the step size for quantizing the low frequency component.