KR100255747B1 - Method and apparatus for encoding a video signal of a contour of an object - Google Patents

Abstract

PURPOSE: A method for encoding a contour video signal of an object and an apparatus thereof are provided to efficiently reduce the transmission volume of data by using a contour motion estimation technique on the basis of a distance between a previous contour and a current contour. CONSTITUTION: Input contour video data which represent positions of contour pixels are provided to a vertex select block(201), a primary sampling block(210), a vertex mapping block(220) and a motion estimation and motion compensation block(280). Current vertex information generated from the vertex select block(201) is provided to a switch(226). A central point of a current contour is calculated on the basis of input contour video data. A central point of a previous contour is calculated on the basis of previous contour video data generated from a memory(270). Prediction contour video data are provided to a subordinate sampling block(230). Motion compensation vertex information is provided to the vertex mapping block(220) and the subordinate sampling block(230).

Description

Contour video signal encoding method and apparatus therefor

The present invention relates to a method and apparatus for encoding contour information of an object represented by a video signal. More particularly, the present invention relates to a method and apparatus for reducing the amount of transmission data using a contour motion estimation technique.

Generally in digital television systems such as video telephony, TELECONFERENCE and high definition television systems, the video line signal of the video frame signal includes a sequence of digital data called pixel values, so that each video frame signal is A significant amount of digital data is needed to define.

However, due to the limited frequency bandwidth available on conventional transmission channels, especially in low-bit-rate video signal encoders such as video telephony and teleconferencing systems, the amount of data is available through various data compression techniques. Should be compressed or reduced.

As is well known in the art, one of the coding techniques for encoding video signals in low transmission coding systems is the so-called "object-oriented analysis and synthesis coding technique" (OBJECT-ORIENTED ANALYSIS-SYNTHESIS CODING TECHNIQUE). In the scheme, the input video signal is divided into a plurality of objects, and three sets of variables for defining the motion, contour, and pixel data of each object are processed through different coding channels.

One example of such an object-oriented analysis and synthesis encoding technique is the so-called Video Experts Committee 4 (MPEG-4), which is a topic-level interactive feature in applications such as low-speed communications, interactive multimedia, and monitors. It is intended to present an audio-video encoding standard to satisfy content-based interactivity, improved coding efficiency and / or versatility (see MPEG-4 Video Verification Model Version 2.0, International Organization for Standardization, ISO /). IEC JTC1 / SC29 / WG11 N1260, March 1996).

According to MPEG-4, an input video image is divided into a plurality of video object planes (VOPs), which correspond to entities of bit streams that can be accessed and handled by a user. A VOP may also be referred to as an object, and the encoder may represent an input video image in VOP units, i.e., because its width and height are represented by a bounding rectangle that is at least an integer multiple of the 16 pixels (macro block size) surrounding each object. Can be processed in units. The VOP includes color information consisting of the luminance component Y and the color components Cr and Cb and contour information represented by a binary mask in one embodiment.

When processing the contour of an object, the contour information becomes important to synthesize and analyze the object shape. A typical coding method for expressing contour information is a chain coding method. However, the chain coding method has a disadvantage in that a large amount of bits are required to express the contour information even if there is no loss of the contour information.

Therefore, in order to solve this problem, various contour information encoding methods such as a polygon approximation method and a B-spline approximation method have been proposed. Polygonal approximation has the problem that the outline is represented schematically. On the other hand, the B-spline method, although the contour is more accurately represented, requires a higher-order polynomial to reduce the approximation error, resulting in an increase in the total amount of computation of the video encoder.

In this approximation method, a contour sine transform (DST) based on a discrete sine transform (DST) is used as a method for eliminating the problems associated with the rough representation of the contour and the increase in the amount of computation.

Multiple approximation as disclosed in US Patent No. 08 / 423,604, co-pending with this patent, entitled “A CONTOUR APPROXIMATION APPARATUS FOR REPRESENTING A CONTOUR OF AN OBJECT” In the apparatus using the method and the contour approximation technique based on the DST, a plurality of vertex points are determined, and the outline of the object is approximated using a polygonal approximation method that maps the contour into line segments. N sample points for each line segment are then selected, and an approximation error is calculated at each of the N sample points to obtain one set of approximation errors for each line segment. The N sample points are equally spaced on each line segment, and each approximation error represents a distance or displacement between each N sample points and the contour. Then, DST coefficient sets can be obtained by doing one-dimensional DST for each set of approximation errors.

Although the outline approximation technique based on the DST described above can reduce the schematic representation and complicated calculations, and thus reduce the amount of data to be transmitted, it would still be desirable to reduce the amount of data to be transmitted.

Therefore, an object of the present invention is to provide an improved contour coding method and apparatus which can effectively reduce the amount of data transmission by using the contour motion estimation technique based on the difference between the front contour and the current contour.

In order to achieve the above object, according to the present invention, the video signal for the current contour of the object is encoded based on the previous contour of the object having a predetermined number of prior vertices (previous vertex) A method of encoding an outline video signal of an object includes: a first process of mapping each vertex to the current contour and generating a prediction vertex corresponding to each vertex on the current contour; Calculating a vertex displacement between each of the vertexes and a corresponding prediction vertex; Determining a plurality of current points on the current contour; A fourth step of generating intra-coded data by encoding the current contour using the current point; A fifth process of generating inter-coded data by encoding the current contour based on the entire contour; And a sixth process of selecting one of the inner encoded data or the inter encoded data based on the vertex displacement as encoded data for the current contour.

1 is a block diagram of an apparatus for encoding input contour image data of an object according to the present invention.

2 is an exemplary diagram for explaining a vertex mapping process according to the present invention.

3 (a) and 3 (b) are schematic diagrams illustrating the main and sub sampling processes according to the present invention, respectively.

* Explanation of symbols for main parts of the drawings

200: contour video signal encoding apparatus 201: vertex selection block

210: main sampling block 220: vertex mapping block

225: mode determination block 226: switch

227: vertex encoder 230: sub-sampling block

235 subtractor 240 transform and quantization block

245 statistical encoder 250 inverse transform and inverse quantization block

255: adder 260: outline playback block

270: memory 280: motion estimation and compensation block

290: Multiplexer (MUX) CC: Hyun Contour

PC: Predictive contours A, B, C, D, E: Motion compensation vertices

A´, B´, C´, D´, E´: prediction peaks

P1, P2, P3, P1´, P2´, P3´: Sampling point

d ₁ ', d ₂ ', d ₃ ': Major error d ₁ , d ₂ , d ₃ : Minor error

Ⅰ ₁ , Ⅰ ₂ , Ⅰ ₃ , Ⅰ ₁ ´, Ⅰ ₂ ´, Ⅰ ₃ ´: intersection

Referring to FIG. 1, a block diagram of an apparatus 200 for encoding input contour image data is shown in accordance with the present invention.

The input contour image data indicating the position of the contour pixels constituting the current contour of the object may include a vertex selection block 201, a primary sampling block 210, a vertex mapping block 220, and motion estimation and A motion estimation and motion compensation (ME & MC) block 280 is supplied. The vertex selection block 201 determines a number of vertices on the current contour using a conventional polygonal approximation technique to generate a plurality of current line segments on the current contour, which are adjacent to each other on the current contour. It is formed by connecting two placed vertices. Vertex information representing the position of the vertex generated in the vertex selection block 201 is supplied to the switch 22.

On the other hand, the motion estimation and compensation block 280 obtains a center point for the current contour and the total contour by averaging the coordinates of all pixel positions with respect to the object outline and the contour, and calculates a motion vector representing the spatial displacement between the two points. Hereinafter, referred to as "Global motion vector (CMV)"). The center point of the current contour is calculated based on the input contour image data, while the center point of the front contour is calculated based on the entire contour image data reproduced in the memory 270, and the total contour image data is the contour constituting the entire contour. The locations of the pixels and their advantages are shown.

Then, the prediction contour is generated by overlapping the entire contour on the current contour. In other words, in the motion estimation and compensation block 280, all pixels on the front contour are moved by GMV to match the center of the front contour with the center of the current contour, thereby generating a predictive contour. The vertexes are also moved by GMV in motion estimation and compensation block 280 to generate motion compensation vertices. In the motion estimation and compensation block 280, the GMV is supplied to the multiplexer (MUX) 290 via the path L20, and the predictive contour image data indicating the position of the contour pixels of the prediction contour is subsampled through the path L30. The motion compensation vertex information indicating the position of the motion compensation vertices is supplied to the vertex mapping block 220 and the subsampling block 230 through the path L40.

In response to the motion compensation vertex information and the input contour image data, vertex mapping block 220 determines a prediction vertex for each motion compensation vertex and calculates a displacement between the two vertices, the prediction vertex being the respective motion compensation vertex. Represents the outline pixel on the nearest outline. Prediction vertex information indicative of the location of the prediction vertices is supplied to the switch 226 via the path L50, and a vertex motion vector representing the displacement between each prediction vertex and the compensating motion vertex corresponding to the prediction vertex is determined by the path L60. Supplied to block 225 and multiplexer 290.

2 illustrates the vertex mapping process performed in vertex mapping block 220, where CC represents the current contour and A to E represent the motion compensation vertices on the prediction contour PC. As shown in the figure, motion compensation vertices A through E are predicted vertices A 'through E', respectively, which are points on the current outline CC closest to the corresponding motion compensation vertex.

Referring back to FIG. 1, the mode determination block 225 determines an encoding mode of the current contour based on the vertex motion vector supplied from the vertex mapping block 220. In particular, to determine the encoding mode, mode determination block 225 calculates the magnitudes of the vertex motion vectors, compares each magnitude with a constant threshold TH, and calculates the number of vertex motion vectors having a magnitude greater than the threshold TH. Calculate

If the number calculated as described above is greater than or equal to a constant integer P, the current outline and the total outline are considered to be considerably different, so that the encoding mode is determined to be intra-mode, so that the current outline is extracted without considering the total outline. Will be encoded. On the other hand, if the calculated number is smaller than the constant integer P, since it is considered that there is considerable similarity between the current contour and the total contour, the coding mode for the current contour is determined as the inter-mode, so that the current contour is equal to the total contour. The liver is encoded based on the difference between the current contours.

The mode determination block 225 supplies the first control signal to the switch 226, the subsampling block 230, and the multiplexer 290 when it is determined to be the inner mode, and otherwise supplies the second control signal.

In the inner mode, in response to the first control signal supplied from the mode determination block 225, the switch 226 selects the current vertex as a vertex for the current contour, and vertex data indicating the position of the selected vertex through the path l10. The main sampling block 210, the vertex encoder 227, and the contour reproduction block 260 are supplied to the vertex data in the inner mode as the current vertex information. On the other hand, the sub-sampling block 230 supplies a sub-error value of 0 to the subtractor 235 and the adder 255. In the vertex encoder 227, vertex data whose vertex data is encoded using a known vertex encoding technique such as an arithmetic encoding technique is generated, and the encoded vertex data is supplied to the multiplexer 290 through a path L70.

In main sampling block 210, the selected vertex, i.e., the current vertex, divides the current contour into a number of main contour segments. Each main contour segment represents a portion of the current contour connecting two adjacent selection vertices with the contour pixels disposed between them, and is approximated by the main line segment connecting the two selection vertices. The main sampling block 210 then selects an integer number of N sample points in a predetermined manner on each main line segment, calculates a major error at each sample point, and sets a major error for each major contour segment. To the subtractor 235.

In a preferred embodiment of the present invention, the integer N sample points selected on the main line segment are arranged at equal intervals from each other. The major error represents the distance from the sample point to the intersection of the straight line and the main contour segment perpendicular to the main line segment, which is a sign indicating the distance to the sample point and the intersection and the relative position of the intersection with respect to the main line segment. It includes.

A negative error of zero in subtractor 235 is subtracted from each set of major errors, supplying a set of difference errors for each major error set to transform and quantization (T & Q) block 240. In my mode, the sub-errors supplied by sub-sampling block 230 are all zero, so the difference errors are equal to the main errors, respectively.

Transform and quantization block 240 transforms and quantizes the difference error set using a series of transform methods such as Discrete Sine Transform (DST) or Discrete Cosine Transform (DCT) to quantize the transform for each difference error. Generate coefficients. Each set of quantized transform coefficients is sent from transform and quantization block 240 to statistical encoder 245 and inverse transform and inverse quantization (IT & IQ) block 250. In statistical encoder 245, each quantized transform coefficient is encoded using known statistical coding techniques, such as variable line length coding (VLC). The coded error data for each set of quantized transform coefficients is supplied to the multiplexer 290 via path L80 at the statistical encoder 245.

In my mode, a first control signal is supplied to the multiplexer 290 via path L90, and in response to this first control signal, multiplexer 290 is encoded via path L80 and coded through path L70. The vertex data is selected and supplied as coded contour data for the current contour to the transmitter for transmission (not shown).

On the other hand, in inverse transform and inverse quantization block 250, each quantization transform coefficient set is converted into a reproduced difference error set and supplied to adder 255, where the reproduced difference error set is converted into a reproduced main error set and The reproduced main error set is supplied to the contour reproduction block 260. In my mode, since the minor errors input at the sub-sampling block 230 are all equals, each reproduced major error is equal to the corresponding reproduced difference error. In the contour reproduction block 260, the current contour is reproduced based on the vertex data supplied to the path L10, and the reproduced main error set and the reproduced current contour image data are supplied to the memory 70, so as to transfer the previous contour to the next contour. Stored as outlines, the reproduced current outline image data includes positional information about vertices and outline pixels of the reproduced current outline.

In the inter mode, the mode decision block 225 supplies a second control signal to the switch 226, the subsampling block 230, and the multiplexer 290 via the path L90, in response to the second control signal. 226 selects the predicted vertex as the vertex with respect to the current contour, supplies vertex data indicating the position of the selected vertex on the path L10, and vertex data in the liver mode is the same as the predicted vertex information.

In main sampling block 210, the current contour is divided into a number of main contour segments by the selected vertices. In the inter mode, each main contour segment represents a portion of two selected vertices adjacent to each other, that is, a current contour connecting two prediction vertices and a contour pixel disposed between them, and is approximated to a main line segment connecting two prediction vertices. Next, the main error set is generated in the same way as in my mode.

On the other hand, in response to the second control signal supplied from the mode decision block 225, the sub-sampling 230 is the same as the method obtained in the main sampling block 210 based on the prediction contour and the motion compensation vertices on the prediction contour. The method determines the minor error set. That is, the prediction contour is divided into a plurality of subcontour segments, each subcontour segment being approximated to a subline segment connecting two motion compensation vertices located at the ends of the main contour segment, and with respect to the main sampling block 210. In the same way as the obtained method, a minor error set representing N displacements between the sub contour segment and the corresponding sub line segment is generated. The minor error set is supplied to subtractor 235 and adder 255.

As shown in FIG. 2, the motion compensation vectors A to E on the prediction contour PC and the prediction vertices A 'to E' on the current contour CC correspond one-to-one to each other, and therefore, each main contour segment, eg For example, the contour C'D 'corresponds to a subcontour segment, for example contour CD. In the subtracter 235 the minor error set for each minor contour segment is subtracted from the major error set for the corresponding major contour segment. For example, as shown in FIGS. 3 (a) and 3 (b), the main errors for the main contour C′D ′ are determined as d ₁ ′, d ₂ ′, d ₃ ′, and the subcontour If the minor errors for the segment are calculated as d ₁ , d ₂ , d ₃ , then d _i (i = 1,2,3) is subtracted from d _i ′ and the difference error c _i (d _i ′ −) in the subtractor 235. d _i ). In FIGS. 3 (a) and 3 (b), N is 3 and I ₁ 'to I ₃ ' and I ₁ to I ₃ are sampling points P ₁ 'to P ₃ ' on the main line segment C'D ', respectively. The intersection point between the straight line drawn at and the sampling points P ₁ to P ₃ drawn on the sub line segment CD and the main contour segment C'D 'and the sub contour segment CD is shown. The difference error set for each major contour segment is transformed and quantized to produce quantized transform coefficients.

Meanwhile, the quantized transform coefficient set is converted into a difference error set reproduced in the inverse transform and inverse quantization block 250 and supplied to the adder 255, where the reproduced error set is added to the corresponding sub-error set and reproduced. An error set is generated and the reproduced main error set is supplied to the contour reproduction block 260. The current outline reproduced in the outline reproduction block 260 is generated based on the vertex data on the path L10 and the reproduced main error set. The reproduced current contour image data representing the reproduced current contour vertex and the positional information about the contour pixel is supplied from the contour reproduction block 260 to the memory 70 and stored to perform a process for the next contour.

In the statistical encoder 245, the quantized transform coefficient set is processed in the same manner as in my mode, and supplies coded error data for the quantized transform coefficient set to the multiplexer 290 via path L80. In inter mode, a second control signal on path L90 is provided to multiplexer 290. In response to the second control signal, the multiplexer 290 sequentially selects coded error data on the path L80, GMV on the path L20, and vertex motion vectors on the path L60, among several signals on the paths L20, L60, L70, and L80, The coded contour data of the current contour is supplied to a transmitter for transmission (not shown). In the decoder at the receiving end, the predicted vertex information, that is, the vertex information supplied from the switch 226 in the inter mode is generated by the GMV, the vertex motion vector, and the vertex information stored in the memory of the decoder, and the outline is reproduced by the outline. Playback is the same as in block 260.

The presently disclosed embodiment is to be considered in all respects as illustrative and not restrictive. It is intended that the present invention be shown by the claims of the invention and embrace all such changes as come within the meaning and range equivalent to the claims.

Claims (13)

A method of encoding a video signal of a current contour of the object based on a previous contour of an object, wherein a predetermined number of prior vertices are provided on the whole contour. The method of claim 1, further comprising: mapping each of the vertexes to the current contour, and generating a prediction vertex corresponding to each of the vertexes on the current contour; Calculating a vertex displacement between each of the vertexes and a corresponding prediction vertex; A third process of determining a plurality of current points on the current contour; A fourth step of generating intra-coded data by encoding the current contour using the current point; A fifth process of generating inter-coded data by encoding the current contour based on the entire contour; And a sixth step of selecting one of the inner coded data and the inter coded data as coded data for the current contour, based on the vertex displacement. The method of claim 1, wherein the first process comprises: an eleventh process of obtaining two center points of the two outlines by averaging positions of the pixels on the whole outline and the current outline; A twelfth step of calculating a center point displacement between the two center points; A thirteenth step of overlapping the entire contour and the current contour by moving one of the two contours toward the other contour by the center point displacement; And a fourteenth process of generating a plurality of prediction vertices corresponding to the plurality of vertexes by finding one pixel on the nested zenith contour closest to each vertex on the nested front contour and determining the pixel as a prediction vertex. Contour video signal encoding method comprising a. The method of claim 1, wherein the fourth process comprises; Approximating the chord contour to a plurality of chord line segments formed by connecting two adjacent vertices on the chord outline, wherein a plurality of chord approximation errors indicative of the difference between the chord contour and the polygonal contour formed by the chord line segments A 41 st process of generating a; Converting the plurality of current approximation errors to generate a plurality of current conversion errors; And a forty-third step of generating inner coded data including the vertex information indicating the position of the vertex and the plurality of current transform errors. 4. The method of claim 3, wherein the fifth process comprises: approximating the current contour to a plurality of main line segments formed by connecting two adjacent prediction vertices on the current contour, and formed by the current contour and the plurality of main line segments. A fifty-first step of generating a plurality of major approximation errors representing the difference between the polygonal contours; A plurality of minor approximation errors representing the difference between the fore contour and the polygonal contour formed by the plurality of sub line segments by approximating the fore contour to a plurality of sub line segments formed by connecting two adjacent vertexes on the fore contour A 52nd process of generating; Obtaining a plurality of difference errors representing a difference between the plurality of major approximation errors and the plurality of minor approximation errors; A 54th step of generating a plurality of difference conversion errors by converting the plurality of difference errors; And a fifty-fifth step of generating inter-coded data including the displacement information representing the vertex displacement, the center point displacement, and the plurality of difference transformation errors. A video signal encoding method for encoding a video signal of a current contour of the object based on a previous contour of an object, wherein the contour video signal encoding method has a predetermined number of prior vertices on the whole contour. The method may include: a first process of mapping each vertex to the current contour and generating a prediction vertex corresponding to each vertex on the current contour; And a second process of encoding the current contour based on the prediction vertex and the vertexes on the contour. 6. The method of claim 5, wherein the first process comprises: an eleventh process of obtaining two center points for the two outlines by averaging positions of the pixels on the whole outline and the current outline; A twelfth step of calculating a center point displacement between the two center points; A thirteenth step of superimposing the front contour and the current contour by moving one of the two contours toward the other contour by the center point displacement; And a fourteenth process of generating a plurality of prediction vertices corresponding to the plurality of vertexes by finding one pixel on the nested zenith contour closest to each vertex on the nested front contour and determining the pixel as a prediction vertex. Contour video signal encoding method comprising a. 7. The method of claim 6, wherein the second process comprises: approximating the current contour to a plurality of main line segments formed by connecting two prediction vertices adjacent to the current contour, by the current contour and the plurality of main line segments. A twenty-first step of generating a plurality of major approximation errors representing the difference between the formed polygonal outlines; A plurality of sub-approximation errors representing the difference between the fore contour and a polygonal contour formed by the plurality of sub-line segments by approximating the fore contour to a plurality of sub line segments formed by connecting two adjacent vertexes on the fore contour A twenty-second process of generating; A twenty-third step of obtaining a plurality of difference errors representing a difference between the plurality of major approximation errors and the plurality of minor approximation errors; A twenty-fourth step of generating the plurality of difference conversion errors by converting the plurality of difference errors; And a twenty-fifth step of generating inter-coded data including the plurality of difference transformation errors. 8. The method of claim 7, further comprising: a third step of calculating a vertex displacement between each of the vertexes on the superimposed all-contour and predicted vertices corresponding thereto; A fourth process of determining a plurality of current points on the current contour; A fifth step of generating an intra-coded data rule by encoding the current contour on the plurality of current points; And a sixth process of selecting one of the inner coded data and the inter coded data as the coded data for the current contour based on the vertex displacement. 9. The method of claim 8, wherein the fifth process comprises: approximating the current contour to a plurality of string line segments formed by connecting two adjacent vertices on the string contour, by the string contour and the plurality of string line segments. A fifty-first step of generating a plurality of chord approximation errors representing a difference between the formed polygonal outlines; A 52 th step of converting the plurality of current approximation errors to generate a plurality of current conversion errors; And a fifty-seventh step of generating inner coded data including the current point information indicating the position of the current point and the plurality of current transformation errors. 8. The method of claim 7, further comprising: a third step of calculating a vertex displacement between a plurality of vertexes on the superimposed entrail and predicted vertices corresponding to the vertexes; A fourth process of comparing the magnitude of each vertex displacement with a constant threshold TH; And a fifth step of selecting the coded data as the coded data for the current contour when the number of vertex displacements whose size is larger than the threshold TH is larger than a constant integer P. A sixty-first step of determining a plurality of vertices on the current contour when the number of vertex displacements is less than P; A sixty-second step of encoding the current contour based on the plurality of current vertices to generate intra-coded data; And a sixth step of generating the inner coded data as coded data for the current contour. A video signal encoding apparatus encoding a video signal of a current contour of the object based on a previous contour of an object, wherein a predetermined number of prior vertices are provided on the whole contour. The apparatus of claim 1, further comprising: vertex determination means for determining a plurality of vertices on the contour; Mapping means for mapping the respective vertices on the current contour and generating a prediction vertex corresponding to the respective vertices on the current contour; Vertex displacement calculation means for calculating a vertex displacement between each of the front vertices and a corresponding prediction vertex; Encoding mode determination means for determining an encoding mode of the current contour based on the current point, and generating a first control signal representing an intra-mode or a second control signal representing an inter-mode. ; Vertex selecting means for selecting the plurality of current vertices and the plurality of prediction vertices as a plurality of selection vertices respectively in response to the first and second control signals; And encoding means for encoding the current contour on the basis of the selection vertex. 12. The apparatus of claim 11, wherein the encoding means comprises: chord outline approximation means for generating a plurality of main approximation errors by approximating the chord contour based on the plurality of selection vertices; Total contour approximation means for generating a plurality of sub-approximation errors by approximating the total contours based on the plurality of selection vertices, wherein the plurality of sub-approximation errors in response to the first control signal is zero; Subtraction means for generating a plurality of difference errors by subtracting the plurality of sub-approximation errors from the plurality of major approximation errors, respectively; Conversion means for converting the plurality of difference errors to generate a plurality of difference conversion errors; And selecting the current point information indicating the positions of the plurality of current points and the plurality of difference transformation errors as encoded data for the current contour in response to the first control signal, and selecting the plurality of difference transformation errors in response to the second response signal. And a selection means for selecting the plurality of difference transformation errors and the displacement information indicating the vertex displacement of the edge. 12. The apparatus of claim 11, wherein the mapping means comprises: averaging means for obtaining two center points for the two contours by averaging positions of pixels on the whole contour and the current contour; Center point displacement calculating means for calculating a center point displacement between the two center points; Moving one of the two contours toward another contour by the center point displacement to find a pixel on the nested current contour closest to each vertebral point on the front contour, and determining the pixel as a predictive vertex And a prediction vertex determining means for generating a plurality of prediction vertices corresponding to the contour video signal encoding apparatus.