KR0181056B1 - Apparatus for encoding difference vectors generated by polygonal approximation of a contour image - Google Patents

Abstract

The present invention relates to an apparatus for encoding an error due to polygonal approximation of a contour image capable of realizing encoding of a high compression rate by extracting an error between a circle outline and an approximated polygon, and vectorizing and quantizing the error. Polygon approximation means for approximating two straight lines into a polygon with a side, wherein a distance between a straight line perpendicular to each side of the approximated polygon and a point meeting a circle outline does not exceed a preset value, and the polygon approximation means of the approximated polygon A plurality of sampling detection means for detecting a vector comprising the error value by detecting the error value between each side of the polygon whose circle outline is approximated and the circle outline with a preset number for each side. And a codebook connected to the plurality of sampling detection means in a one-to-one correspondence. And a plurality of optimal pattern selection means for detecting the index of the pattern that is optimal for the pattern of the error vector of each side, and encoding means for variable length encoding the indices of the plurality of optimal pattern selection means.

Description

Error Coding Apparatus Due to Polygonal Approximation of Contour Image

1 is a diagram for explaining a general polygon approximation.

Figure 2a is an overall configuration diagram showing an embodiment of the present invention.

FIG. 2B is a detailed configuration diagram illustrating the polygon approximation part of FIG. 2A in more detail. FIG.

FIG. 2C is a detailed configuration diagram showing the sampling detection unit of FIG. 2A in more detail. FIG.

3A is a diagram for explaining the operation of the polygon approximation unit in FIG. 2B.

3b, c, and d are diagrams for explaining a process of vector quantizing an error due to polygonal approximation of a contour image.

* Explanation of symbols for main parts of the drawings

21: contour information detector 22: polygon approximation unit

23 to 26: sampling detection unit 27 to 30: optimal pattern selection unit

31: encoder

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, encoding by vector quantizing an error due to polygonal approximation of an outline image. It relates to an encoding device.

Recently, studies on video encoding techniques of low transmission that can be accommodated ahead of MPEG (Motion Picture Experts Group) 4 have been actively conducted.In order to encode low transmission of continuous images including moving objects, data using moving information is used. Reduction techniques are 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, image quality degradation such as blocking, corner blur, and spots may appear in the playback image by applying block-based motion compensation coding.

Therefore, in MPEG4, unlike the conventional method which considers only the statistical characteristics of the video signal, the video encoding technique that focuses on the subjective image quality based on human visual characteristics and can significantly reduce the amount of transmission data is studied. One of the new coding schemes is proposed as an object-based coding scheme that compensates motion for each object by dividing a video into moving objects and estimating motion expression parameters for each moving object.

In the object-specific encoding technique, first, contour information, motion information, and color information are extracted by dividing an input image into moving objects of independent motion and estimating the motion information of the divided object. , Motion information, and color information are selectively encoded and transmitted for each moving object.

Since most regions can be motion compensated by motion coefficients, only the motion information and the contour information need to be transmitted. Accordingly, it is known that more data reduction effects can be obtained than the block coding method.

The most widely known outline coding method is chain coding, but chain coding, which is lossless coding, has a large amount of data and is not suitable for ultra low speed video coding.

Therefore, as a lossy coding technique that can reduce a lot of data, a method of approximating a circular outline to a polygon connected with a straight line is generally known. As suggested by tter (Object Oriented Analysis Synthesis Coding based on Moving Two-Dimensional Objects, Signal Processing: Image Communication Vol. 2, pp.409-428, 1990), polygon approximation can be used to create a sequence of contour images. It is approximated by spline), and the error is encoded by the motion compensation prediction of the vertex of the polygon and spline appearing at this time.

Referring to FIG. 1, polygonal approximation is a method of forming a straight line (b) based on a vertex (a) on a contour image to express the outline as a plurality of straight lines, wherein the actual contour and the approximate polygon are used as a criterion for determining the degree of approximation. The maximum distance d <d _max is used and the maximum distance d _max determines the performance of the polygon approximation. However, in the contour image obtained by polygon approximation, a rough reconstructed image is obtained even when the maximum distance d _max is relatively small.

In other words, polygonal approximation can encode the contours with a relatively small amount of data, but the contours are roughly expressed, which is not only very unnatural to human vision, but also results in a lot of distorted circular contours. have.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned problems. An apparatus for encoding an error due to polygonal approximation of a contour image capable of realizing encoding of a high compression ratio by extracting an error between a circular outline and an approximated polygon and vectorizing the error is provided. The purpose is to provide.

In order to achieve the above object, the present invention approximates a circle outline of an image to a polygon with a plurality of straight lines, but the distance between a point perpendicular to a straight line and a circle outline perpendicular to each side of the approximated polygon exceeds a preset value. Polygon approximation means for approximating polygons that do not approximate polygons, and the number corresponding to each side of the approximated polygon, and detecting the error value between each side of the polygon to which the circle outline is approximated and the circle outline with a preset number for each side. A plurality of sampling detection means for detecting a vector having an error value as a component, and a plurality of sampling detection means connected to the plurality of sampling detection means in one-to-one correspondence and detecting an index of a pattern that is optimal for a pattern of an error vector of each side by using a preset codebook Variable pattern coding means for each of the plurality of optimal pattern selection means and the indexes of the plurality of optimal pattern selection means It is characterized in that it consists of encoding means.

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

2A is a detailed configuration diagram for implementing the present invention. The contour information detection unit 21 divides an input image into moving objects of independent motion and detects contour information through an image analysis process of estimating motion information of the divided object. By connecting the straight lines based on the vertices of the circular contours of the image detected by the contour information detector 21, the contours are approximated to a plurality of straight lines to form a polygon with a plurality of straight lines on the sides. Polygon approximation section 22 that forms a polygon whose distance (d) between a straight line perpendicular to each side and a point that meets a circle contour does not exceed a preset value (d _max ), and the number corresponding to each side of the polygon The sample which detects the error value between each side of the polygon whose circle outline is approximated and the circle outline by the preset number for each side and detects the vector whose error value is a component One-to-one correspondence to the detectors 23, 24, 25, and 26 and the sampling detectors 23, 24, 25, and 26. The index of the pattern that is optimal for the pattern of the error vector of each side using a preset code book. Is composed of an encoding pattern 31 for variable length encoding the respective indexes of the optimum pattern selection section 27, 28, 29, 30 for detecting the P-values, and the indexes of the optimum pattern selection section 27, 28, 29, 30.

In this case, as shown in FIG. 2B, the polygon approximation unit 22 selects N-1 (N; preset positive integer values) that select and output the coordinates of the previously detected contour information and the coordinates of the newly detected contour information. N node memories 311, 312, 313, 314 including selectors 301, 302, 303, 304 and two node memories 311, 312 that store the coordinates of one end point and the other end point of the circular contour information, respectively. , 315, and N-1 selectors 301 and 302 at input terminals of the remaining node memories 312, 313, 314, and 315 except for the node memory 311 storing one end point coordinates of the circular contour information. Node memory groups 311, 312, and 313 connected to one-to-one correspondences of the output terminals and storing coordinate values of the contour information selected by the selection units 301, 302, 303, and 304 in the corresponding node memories. 314, 315, and preselected contours stored in the node memory groups 311, 312, 313, 314, and 315, respectively. N-1 error detectors 321, 322, 323, and 324 that detect the maximum distance between the line perpendicular to the straight line connecting the adjacent coordinates and the point where the contour meets, among the coordinates of N, N-1, which is connected in a one-to-one correspondence to 321, 322, 323, and 324 and receives adjacent coordinate values among the coordinates of the preselected contour information, respectively, and detects the contour coordinates of the maximum distance detected by the error detector connected in one-to-one correspondence. Comparing the maximum distances detected by the two coordinate detectors 331, 332, 333, 334 and the N-1 error detectors 321, 322, 323, and 324 with each other to select the largest value (maximum value). value selection unit 341, the maximum selected by the maximum value selection unit group the maximum value selected by the (341) set value (d _max) and compared to a preset value (d _max) selected is greater than the maximum value unit 341 Selection that allows you to select the coordinates of the point where the value meets the contour information. The maximum value selector 341 of the coordinates output from the N-1 coordinate detectors 331, 332, 333, and 334 by the selection control signal of the comparator 351 and the comparator 351 for generating the control signal. I, i + 1 th (where the two coordinate values adjacent to the coordinate value newly detected by the coordinate selecting unit 361 and the coordinate selecting unit 361 respectively detect the coordinates of the point where the maximum value selected by the contour information meets the contour information). iN) Controls the i + 1th selector so that newly detected coordinates are stored in the i + 1th (i + 1≤N) node memory in which the lower coordinates are horizontally stored in the node memory, except for the i + 1th selector. Each selector selects coordinate values outputted from the node memory of the previous stage, and the new coordinates are selected by the selection controller 371 and the coordinate selector 361 that control the shift to the next node memory. When inputted to the selector 301, 302, 303, 304, all of the i + 1th node memories Node memory in the block is composed of a DJ (disable) and the memory controller 381 to the enable (enable) to the rear end node memory including the node (i + 1) th memory.

Each of the sampling detectors 23, 24, 25, and 26 is one of coordinates of pre-selected contour information stored in N node memory groups 311, 312, 313, 314, and 315, respectively, as shown in FIG. A length detector 411 which receives the adjacent coordinate values and detects the length of the vertical line connecting the adjacent coordinates, that is, the length 1 for each side of the polygon whose circle outline is approximated, and the preset number of each side of the polygon ( The distance from the equalization point of each side divided by the division part 412 and the division part 412 to the point which meets the straight line perpendicular | vertical to each side and a circle contour (divided by k + 1) A quantization unit 413 for quantizing the error of &lt; RTI ID = 0.0 &gt;

An operation description of an embodiment of the present invention configured as described above will be described with reference to FIGS. 3a, b, c, and d.

First, the contour information detector 21 divides the input image into moving objects of independent motion and estimates the contour information, motion information, and color information among the contour information, motion information, and color information detected through an image analysis process. Coordinates of one end point and the other end point of the information are extracted and output to the polygon approximation unit 22.

Coordinates of one end point A and the other end point B of the contour information extracted by the contour information detector 21 are stored in the node memory 311 and the node memory 312, respectively, and then input to the error detector 321. The error detector 321 receives the contour information of the contour information detector 21 and the coordinates of one and the other end points of the contour information, respectively, and as shown in FIG. 3A, a straight line and a contour perpendicular to a straight line connecting one and the other endpoint coordinates. The distance (d) is the maximum distance to the point that meets the point and the coordinate detection unit 331 connected to correspond to the error detector 321 is a point that meets the straight line and the contour line perpendicular to the straight line between the one side and the other end point coordinates The outline coordinate C of the distance d1 having the maximum distance to is detected and output to the coordinate selecting unit 361.

The maximum value selector 341 compares the maximum distances detected by the N-1 error detectors 321, 322, 323, and 324 with each other and selects the largest value (maximum value). Since it is output only at 321, the maximum distance d1 output from the error detector 321 is selected and output to the comparator 351, and the comparator 351 uses the maximum value selected by the maximum value selector 341. set value and generates a selection control signal to be selected coordinate of the point selected maximum value is met and the outline information by the (d _max) and compared to a preset value (d _max) selected is greater than the maximum value unit 341.

Therefore, the coordinate selector 361 receives each coordinate of the contour information output from the N-1 coordinate detectors 331, 332, 333, and 334 and selects one coordinate according to the selection control signal of the comparator 351. The input unit is input to each of the N-1 selectors 301, 302, 303, and 304. Initially, only the C point coordinate values detected by the coordinate detector 331 are output, so that the coordinate selector 361 coordinates the coordinates. The C point coordinate values output from the detector 331 are selected to be input to the N-1 selectors 301, 302, 303, and 304, respectively.

In this case, when the newly detected coordinate values are input to the N-1 selectors 301, 302, 303, and 304, respectively, the selection controller 371 stores node coordinates (i-th node) that store the coordinate values adjacent to the newly detected coordinate values. Memory and the i + 1 th node memory) and then control a selector connected to correspond to the i + 1 th node memory in which the lower coordinates are horizontally stored among the node memories in which adjacent coordinate values are stored. The remaining selector controls to select the coordinate values output from the node memory of the previous stage and shift them to the node memory of the next stage.

Accordingly, the C point coordinate value is stored in the node memory 312 by the selection of the selection unit 301 and the B point coordinate value stored in the original node memory 312 is the next node memory by the selection of the selection unit 302. 313 is stored.

By the above process, among A, C, and B point coordinate values stored in each node memory 311, 312, and 313, adjacent A and C point coordinate values are input to the error detector 321, and C The point and B point coordinate values are input to the error detector 322.

The error detector 321 detects a distance d2 having a maximum distance between a straight line perpendicular to a straight line connecting the point A and C coordinates and the point where the contour meets the outline, and is connected to the error detector 321. 331 detects contour coordinates D of a distance d2 having a maximum distance between a straight line perpendicular to a straight line connecting A and C coordinates and a point meeting an outline, and outputs it to the coordinate selecting unit 361. do.

In addition, the error detector 322 detects a distance d3 having a maximum distance from a straight line perpendicular to a straight line connecting the C point and B point coordinates and the point meeting the contour line, and coordinates connected corresponding to the error detector 321. The detector 332 detects the contour coordinate E of the distance d3 having the maximum distance from the straight line perpendicular to the straight line connecting the C point and the B point coordinates and the point meeting the contour line, to the coordinate selector 361. Output

The maximum value selector 341 compares d2 and d3 and selects a larger value. For example, if d3 is larger, the maximum value selector 341 outputs it to the comparator 351, and the comparator 351 d is a preset value d. _{When max} is compared and d3 is larger than d _max , the selection control signal is output to the coordinate selection unit 361.

Therefore, the coordinate selector 361 selects the coordinate values of the point E detected by the coordinate detector 332 and inputs the coordinate values to the selectors 301, 302, 303, and 304, respectively.

In this case, when the newly detected E point coordinate values are input to the N-1 selectors 301, 302, 303, and 304, respectively, the selection controller 371 receives the C point and B point coordinate values adjacent to the newly detected coordinate values, respectively. After detecting the stored node memories 312 and 313, the newly detected E point coordinates are controlled by controlling the selector 302 connected to correspond to the node memory 313 in which the lower coordinates are horizontally stored among the node memories in which adjacent coordinate values are stored. The values are stored in the node memory 313, and the remaining selector controls to select the coordinate values output from the node memory of the previous stage and shift them to the next node memory.

On the other hand, when the E point coordinate value is selected by the selection unit 302, the memory control unit 381 deactivates the node memory 312 located at the front of the node memory 313 and includes the node at the rear end including the node memory 313. Since the memories 314 and 315 are enabled, when the N-1 selectors 301, 302, 303 and 304 are operated, the coordinate values of all states remain in the node memory 312 and the node memories 314 and 315 are operated. 315 is newly stored the coordinate values of the previous state that was stored in the node memory of the front end.

By the above process, among the A, C, E, and B point coordinate values stored in each of the node memories 311, 312, 313, and 314, the A and C point coordinate values adjacent to each other are detected by the error detector 321. The C and E point coordinate values are input to the error detector 322 and the E and B points are input to the error detector 323.

The error detector 321 detects a distance d2 having a maximum distance between a straight line perpendicular to a straight line connecting the point A and C coordinates and the point where the contour meets the outline, and is connected to the error detector 321. 331 detects contour coordinates D of a distance d2 having a maximum distance between a straight line perpendicular to a straight line connecting A and C coordinates and a point meeting an outline, and outputs it to the coordinate selecting unit 361. do.

In addition, the error detector 322 detects a distance d4 having a maximum distance between a straight line perpendicular to a straight line connecting the C point and the E point coordinates and a point meeting the outline, and coordinates connected corresponding to the error detector 322. The detection unit 332 detects the contour coordinate F of the distance d4 having the maximum distance from the straight line perpendicular to the straight line connecting the C point and the E point coordinates to the point where the contour meets the contour line. Output

Similarly, the error detector 323 detects a distance d5 having a maximum distance between a straight line perpendicular to a straight line connecting the coordinates of the E point and the B point and the point meeting the contour line, and is connected to the error detector 323. The coordinate detector 333 detects the contour coordinate G of a distance d5 having a maximum distance between a straight line perpendicular to a straight line connecting the point E and B coordinates and a point meeting the outline, and then selects the coordinate selector 361. Output to

The maximum value selector 341 compares d2 with d4 and d5 and selects the largest value. For example, if d2 is the largest, the maximum value selector 341 outputs it to the comparator 351, and the comparator 351 outputs a preset value of d2. a, and outputs a selection control signal d2 and d _max is greater coordinates selection section 361 than compared to d _max.

Therefore, the coordinate selecting unit 361 selects the coordinate values of the point D detected by the coordinate detecting unit 332 and inputs them to the respective selecting units 301, 302, 303, and 304.

In this case, when the newly detected D point coordinate values are input to the N-1 selection units 301, 302, 303, and 304, respectively, the selection controller 371 receives the A and C point coordinate values adjacent to the newly detected coordinate values. After detecting the stored node memories 312 and 313, the newly detected D point coordinates are controlled by controlling the selection unit 301 connected to correspond to the node memory 312 in which the lower coordinates are horizontally stored among the node memories in which adjacent coordinate values are stored. The values are stored in the node memory 312, and the remaining selector controls to select the coordinate values output from the node memory of the previous stage and shift them to the next node memory.

On the other hand, when the D point coordinate value is selected by the selection unit 301, the memory control unit 381 enables the node memories 313, 314, and 315 including the node memory 312, so that N-1 selections are made. The coordinates of the D point are stored in the node memory 312 by the units 301, 302, 303, and 304, and the coordinate values of the previous state stored in the node memory of the front end are newly stored in the node memories 313, 314, and 315. Stored.

By repeating the above process, N contour coordinates are stored in the N node memory groups 311, 312, 313, 314, and 315, and the circle outline is stored in the node memory groups 311, 312, 313, 314, and 315. It is approximated by a polygon with the vertex of each stored contour coordinate.

At this time, d _max depends on the performance of polygon approximation, which can be set through experiments, and it will be understood by those skilled in the art that N is determined in consideration of the number of contour coordinates that can be detected according to the set d _max .

Each of the N-1 sampling detection units 23, 24, 25, and 26 respectively receives adjacent coordinate values among coordinates of preselected contour information stored in the N node memory groups 311, 312, 313, 314, and 315, respectively. .

For example, A point coordinate values are stored in the node memory 311, D point coordinate values are stored in the node memory 312, C point coordinate values are stored in the node memory 313, and E is stored in the node memory 314. When the point coordinate values are stored, the A and D point coordinate values are input to the sampling detector 23, the D and C point values are input to the sampling detector 24, and the C and E points are input to the sampling detector 25. The coordinate value is entered.

Accordingly, the length detector 411 of the sampling detector 23 receives coordinates A and D, detects the distance 1 between the two coordinates, and outputs the coordinates 1 to the divider 412.

The division unit 412 calculates an equal distance e to divide the distance 1 between the A-point and D-point coordinates by a predetermined number k + 1, and an expression for this may be expressed as in the first equation. .

For example, as shown in FIG. 3B, the distance between A and D points If the distance between the point A and the point D is to be divided into four (k + 1 = 4), the equal distance (e) is given by (Formula 2) by (Formula 1).

Therefore, the number of equal points (k) of the distance between the point A and the point D is three by the dividing unit 412, and the equal points of the sides are (Figure 3c).

Therefore, the quantization unit 413 A straight line perpendicular to the circle contour between A and D coordinates is formed at the equinox of, and the distance between the point where the vertical line meets the contour and the equidistant point, that is, the error between the circle contour and the polygonal equinox Are respectively detected, and the detected error value is quantized to a predetermined quantization level to detect an error vector (FIG. 3D).

Similarly, the sampling detector 24 inputs the coordinates of the D and C points, detects the distance (1) between the D and C point coordinates, divides them by a preset number (k + 1), and then divides them at equal points of each side An error vector is detected by quantizing a distance between a straight line perpendicular to each side and a point where an outline of a circle meets at a predetermined quantization level.

In the same manner as described above, each of the sampling detectors 23, 24, 25, and 26 receives a polygonal approximation by receiving adjacent coordinate values among the coordinate values of the contour information stored in the node memories 311, 312, 313, 314, and 315. Detect the error vector of the sides.

The optimal pattern selectors 27, 28, 29, and 30 that are connected in one-to-one correspondence with the sampling detectors 23, 24, 25, and 26 are optimal patterns for the pattern of the error vector of each side using a preset code book. The index is detected and output, and the encoder 31 variable-length encodes the index output from each optimum pattern selector 27, 28, 29, or 30.

So far, polygonal approximation of the circle contour has been described with respect to the vector quantization using the error value of each side of the polygon. In this case, two vector quantizations can be presented.

In other words, if each side of the approximated polygon is divided into k + 1 to detect k errors with the circle outline for each side, one k-dimensional vector is obtained for each side, and if the number of polygon sides is N, k N vectors of the difference are detected, and in this case, the preset codebook represents a codebook corresponding to an error vector having only an error value between a circle outline and an approximated polygon.

On the other hand, the longer the length of the straight line, the more components of the error vector should be obtained, which means that the dimension of the error vector increases. Therefore, we propose a method of generating variable vectors of different dimensions according to the length of a straight line and then performing variable vector quantization according to the length of the straight line.

That is, vector quantization is performed by generating a plurality of codebooks of different dimensions according to the distribution of the linear length, and variable vector quantization is performed by different quantization levels.

As described above, the present invention has an effect of realizing a high compression ratio by extracting an error between a circular outline and an approximated polygon and vectorizing and quantizing the error.

Claims (6)

The circle outline of the image is approximated by a polygon with a plurality of straight lines, but the distance (d) between the line perpendicular to each side of the approximated polygon and the point where the circle outline meets does not exceed the preset value (d _max ). A polygon approximation unit 22 for approximating the polygon; Comprising a number corresponding to each side of the approximated polygon (N-1, N: a predetermined positive integer value), and the error value between each side of the polygon to which the circle outline is approximated and the circle outline N-1 sampling detectors 23, 24, 25, and 26 for detecting a predetermined number each time and detecting a vector having an error value as a component; N-1 optimal patterns connected to the N-1 sampling detectors 23, 24, 25, and 26 in a one-to-one correspondence to detect an index of a pattern that is optimal for the pattern of the error vector of each side by using a preset codebook Selector 27, 28, 29, 30; And an encoding unit (31) for variable length encoding the indices of the N-1 optimal pattern selection units (27, 28, 29, 30). 2. The polygon approximation unit 22 according to claim 1, wherein: N-1 selectors 301, 302, 303 and 304 for selecting and outputting the coordinates of the previously detected contour information and the coordinates of the newly detected contour information. ; N node memories 311, 312, 313, 314, and 315 including two node memories 311 and 312 respectively storing coordinates of one end point and the other end point of the circle contour information, and the circle contour information. An output terminal of each of the N-1 selectors 301, 302, 303, and 304 is input to an input terminal of each of the remaining node memories 312, 313, 314, and 315 except for the node memory 311 storing the coordinates of one end of the node. Node memory groups 311, 312, 313, 314, 315 connected one-to-one correspondingly and storing respective coordinate values of the contour information selected by the N-1 selectors 301, 302, 303, and 304 in the corresponding node memories. ); Detecting a distance having a maximum distance between a straight line perpendicular to a straight line connecting adjacent coordinates and a point meeting the contour among coordinates of the previously selected contour information stored in the node memory groups 311, 312, 313, 314, and 315, respectively. N-1 error detection units 321, 322, 323, and 324; The maximum distance detected by the error detection unit connected in one-to-one correspondence to the N-1 error detection units 321, 322, 323, and 324, and receiving adjacent coordinate values among the coordinates of the previously selected contour information. N-1 coordinate detection units 331, 332, 333, and 334 for detecting the contour coordinates of the coordinates; A maximum value selector 341 which selects the largest value (maximum value) by comparing the maximum distances detected by the N-1 error detectors 321, 322, 323, and 324 with each other; The maximum value selected by the maximum value selection unit 341, a preset value (d _max) and compared to the predetermined value (d _max) is greater than the maximum value selection unit 341, a maximum value of the outline information selected by the A comparator 351 for generating a selection control signal for selecting a coordinate of a point meeting with; The maximum value selected by the maximum value selector 341 among the coordinates output from the N-1 coordinate detectors 331, 332, 333, and 334 by the selection control signal of the comparator 351 is determined by the contour information. A coordinate selecting unit 361 which detects coordinates of meeting points; I, i + 1 th (i: positive integer) th i (i: positive integer) node memory where the two coordinate values adjacent to the newly detected coordinate value by the coordinate selecting unit 361 are stored, i The selector corresponding to the i + 1 th node memory is stored so that newly detected coordinates are stored in the +1 th (i + 1≤N) node memory, and the remaining selector except for the selector corresponding to the i + 1 th node memory is selected. A selection control section 371 which controls the N-1 selection sections so as to select coordinate values output from the node memory of the previous stage and shift them to the node memory of the next stage; When a new coordinate is selected by the coordinate selector 361 and input to each of the N-1 selectors 301, 302, 303, and 304, the node memory located in front of the i + 1 th node memory is disabled. And a memory controller (381) for enabling a rear node memory including an i + 1 th node memory. The N-1 sampling detectors 23, 24, 25, and 26 are each one of: coordinates of preselected contour information stored in the node memory group 311, 312, 313, 314, and 315, respectively. A length detector 411 which receives adjacent coordinate values, respectively, and detects a length 1 of vertical lines connecting adjacent coordinates; The division part 412 which equally divides the length 1 of the vertical line detected by the length detection part 411 by the predetermined number k + 1, and the straight line which is perpendicular | vertical at the equal part divided by the said division part 412 And a quantization unit (413) for detecting an error vector by quantizing a distance between the point and the point of contact with the circle contour at a predetermined quantization level, wherein the error is due to the polygon approximation of the contour image. 4. The apparatus of claim 3, wherein the preset codebook comprises a codebook corresponding to an error vector having an error value between a circle outline and an approximated polygon. . The codebook of claim 3, wherein the preset codebook comprises: a codebook corresponding to an error vector having an error value between a circle outline and an approximated polygon, and having a plurality of codebooks having different dimensions according to a distribution of straight lines. An apparatus for encoding errors due to polygonal approximation of contour images. The method according to claim 5, wherein the quantization unit 413 is configured to quantize different quantization levels according to lengths of linear lengths when the preset codebooks are composed of a plurality of codebooks having different dimensions according to the distribution of linear lengths. An apparatus for encoding errors due to polygonal approximation of contour images.