KR890003319B1 - Apparatus for processing character or pictorial image data - Google Patents

Abstract

A method of processing character data includes approximating the outline of a character with a set of functional curves or straight lines, then storing the outline specifying information to compress the amount of data, and decoding the compressed data to reproduce the character image. Sample points on the outline are selected so that the vector joining adjacent points deviates from the outline by less than a predetermined amount. Vectors having greater than a predetermined length are deemed to correspond to straight lines and are encoded as such vectors shorter than the predetermined length correspond to curved portions and are encoded by segmental polynomial approximation.

Description

Text Image Data Processing System

1 and 2 illustrate a conventional outline method.

3 is a flowchart showing an outline of a processing procedure of the system of the present invention.

4 is a view showing an outline of the configuration in which the present invention system is implemented.

5 is a view showing an example of a character outline.

FIG. 6 is a diagram showing an example in which the outline of the character shown in FIG. 5 is divided into a plurality of blocks. FIG.

7 is a diagram for explaining straight line approximation.

8 is a block diagram showing an embodiment of a configuration for realizing identification of straight portions and curved portions.

FIG. 9 is a view showing details of main parts of the process in the configuration of FIG. 8; FIG.

10 is a view showing a result of identifying a straight portion and a curved portion.

11 (a) to 11 (c) are diagrams for explaining a method of calculating the inclination at each contour point.

12 is a diagram illustrating a process of setting a sample section.

13 (a) and 13 (b) are diagrams for explaining a method for determining sample points.

14 is a view showing a situation in which sample points are set on the character outline.

FIG. 15 shows an example of a format for storing arbitrary one block data. FIG.

Fig. 16 is a diagram explaining a storage method of block data.

FIG. 17 is a block diagram showing an embodiment of a configuration for restoring block data. FIG.

18 is a timing chart showing the operation of FIG.

* Explanation of symbols for main parts of the drawings

1, 50, 90: outline of characters 60: connection point storage unit

61: straight part identification vector length setting part 62: curve dividing point identification angle setting part

63: difference connection point coordinate register 64: current connection point coordinate register

65: all connection point coordinate register 66: vector length calculation unit

67: angle calculation unit between the vectors 68: vector length comparison unit

69: angle comparison unit 70: linear storage unit

71: curve storage unit 91: approximation curve of the n-th order polynomial

120: compressed data storage section 121: selector 1

122: magnification memory unit 123: decoder group

124: selector 125: 1-character memory

126: output device

The present invention relates to a character image data processing system which compresses the amount of stored data by storing the outline shape of a character or image (hereinafter referred to as a character). Specifically, the outline of a character is classified into a straight portion and a curved portion. A character image data processing system in which a sample point candidate provided on this curved portion is moved by trial and error to decode the compressed data and reproduce a character image. The determination of whether or not this trial and error is carried out and in which direction the sample point candidate is to be started is determined depending on the amount of deviation? Obtained for each contour point in the sample candidate section. The sample sections are sequentially set at the starting point of the curved section, and the curved section is divided into a plurality of sample sections, and the straight section and each sample section obtained here are stored.

It is well known that two-value data obtained by dot decomposition of characters is extremely redundant. Thus, various data compression schemes have been proposed in order to alleviate this redundancy.

Another one is a data compression method called so-called contour method, in which the shape of a character is identified as an outline and the data amount is specified by storing the data specifying the outline.

As the compression method by this contour method, the linear (vector) approximation method like FIG. 1 and the n-th order curve approximation method like FIG. 2 are already proposed.

The linear approximation method illustrated in FIG. 1 is a method disclosed in JP-A-54-149522, JP-A-55-79154, and the like. The outline is approximated by a set of vectors (2) in which the outline (1) of any character shown by a dotted line is shown by a solid line and specifies each vector (viewpoint position, length and inclination, or horizontal and vertical components). It is possible to compress the data by using as the sign and the data.

The n-th order curve approximation illustrated in FIG. 2 is a method known from the applicant's JP No. 57-39963. As will be described later, an n-th order curve that attempts to compress a data amount by storing coordinates of a point P group appropriately set on the outline of an arbitrary character, and forms an arbitrary (n + 1) point in succession ( It is approximating the desired outline in 3). Figure 2 also illustrates the case of n = 2.

The data compression method by the outline method is characterized by reproducing the compressed data and reproducing the character image by performing interpolation processing, extraction processing or vector magnification conversion processing to reproduce the character image at various magnifications.

However, these conventional methods, for example, the angle of inclination of each left and right tangent line centered on the connection point Pc of the start point, the end point P of each vector in FIG. 1, or the n-th order curve 3 in FIG. As apparent from the fact that δ becomes discontinuous in any case, there was an inherent defect that the smoothness of the contour (the continuity of the inclination of the outline) is not guaranteed.

On the other hand, in general, the outline of a character has both a straight line and a curved line, and the outline itself is not only continuous but also except for the peculiar point such as the intersection of the character image or the tip of the 'beep'. It has the characteristic that the first derivative (the slope of the contour) changes continuously.

Therefore, the conventional data compression method has a problem that it is difficult not only to obtain specific compressed data faithfully to the outline but also to accurately remove the unnaturalness (slope discontinuity) of characters reproduced on the basis of the data.

In order to solve such a problem, the present applicant has already proposed the method of JP-A-58-134745.

The data compression method disclosed herein is also a kind of contouring method, and has the following problems. That is, since (1) all the sample intervals of any one block are approximated at once, in the part where the linear part and the curved part connect, it is easy to generate a deviation from the contour at the connection point, and the part to be reproduced as a straight line also Since the approximation is affected by the approximate curve of the front and rear sample sections, the reproducibility of the straight line is inferior.

(2) Moreover, in order to avoid this, it has become clear that a large number of sample points must be set, divided into many polynomials, and approximated, thereby increasing the amount of data.

(3) In addition, since each sample section in any one block was approximated at once, the problem that the approximation curve of other sample sections also changed when the sample point was newly established was also undone.

(4) In addition, each time a sample point is established, an approximation curve has to be recalculated for all sample intervals on any one block, and furthermore, it takes a long time to prepare the data to be obtained because complex calculation is required for encoding.

SUMMARY OF THE INVENTION An object of the present invention is to provide a data processing system by an improved contour method, and another object of the present invention is to reduce the computation time in obtaining compressed data and to obtain compressed data that faithfully reproduces the contour. It is another object of the present invention to provide a character image data processing system capable of storing compressed data faithfully reproducing the smoothness of outlines and attaining a sufficiently high data compression ratio.

An outline of the processing procedure of the system of the present invention will be described according to the flowchart shown in FIG.

The input character image is dot-decomposed into x and y matrices (30), and the outline of this dot-decomposed character is extracted (31). The extracted contour is linearly approximated with a vector whose deviation from the contour is less than the tolerance. (32) The linearly approximated contour is a vector length and thus identified as a straight line and a curved line (33). If necessary, the curved portion and the identified contour are divided based on the intersection angle of adjacent vectors (34). In preparation for curve approximation of the contour of the curved portion in the cubic polynomial, the inclination at each contour point forming the contour is obtained (35). The third-order polynomial is obtained from the start point, the end point coordinates, and the slope of the sample candidate section set in the curve approximation section. Then, while extending or shortening the sample candidate section, the sample section that is the longest within the range where the deviation between the third-order polynomial and the contour approximating this section is accommodated below the tolerance is obtained, and the start point and the end point of the section are sampled. The decision is made based on the point (36).

L일때, 상기 윤곽 구간[Pi, Pi+1]을 곡선부로 식별하고, 다음의 벡터에 대해서도 반복하여 상기 비교를 행하고 벡터 Vi+k에서 li+k＞L가 될때 또는 Qi+k＜Q가 될때, 그때까지의 윤곽구간[Pi, Pi+k]을 하나의 연속한 곡선부로 식별하는 식별수단, (46)은 상기 곡선부의 윤곽구간 [Pi, Pi+k]에 대하여 윤곽의 시점 Qj를 제1의 샘플점으로 하고 임의의 윤곽점 Qj+r를 샘플점 후보로 하는 샘플후보구간[Qj, Qj+r]을 설정하고 이 샘플점 Qj와 샘플점 후보 Qj+r의 좌표(xj, yj), (xj+r, yj+r)및 그 경사 tj, tj+r에 의거하여 상기 샘플 후보구간 [Qj, Qj+r]을 3차 다항식 f(x)으로 곡선 근사하는 곡선 근사 수단, (47)은 이 3차 다항식 f(x)과 상기 샘플 후보 구간의 각 윤과점과의 편위량 ε을 구하는 편위량 ε산출수단. (48)은 구해진 편위량ε과 허용오차△를 비교하는 제3의 비교수단, (49)는 상기 비교 결과에 따라 현재의 샘플점 후보를 이동시키고, 갱신된 샘플후보구간에 대하여 상기와 같이 평위량 ε와 허용오차△와의 비교를 행하고, 그 비교결과가 모든 윤곽점에 대하여 ε

△을 만족시키고, 또 이 샘플 후보구간이 최대가 되는 경우의 샘플점 후보를 다음의 샘플점으로하여 설정하는 동시에 이 새로 설정한 샘플점을 다음의 샘플 후보 구간의 시점으로하여, 상기 하나의 연속된 곡선부와 식별 한 윤곽구간 [Pi, Pi+k]의 전부에 걸쳐 샘플 구간을 차례로 설정해 가는 샘플 구간 설정수단.In this way, when the sample intervals are obtained in order, and a third order polynomial approximating each sample interval is obtained, first order functions of the respective third order polynomials are obtained, and the slope of each contour is calculated again (37). The sample point is determined as described above according to the slopes and coordinates of each newly obtained contour point, and a third order polynomial faithful to the contour is calculated (38). The view coordinates of the linear sample section obtained as described above and the coordinates of each curve sample section constituting the curved portion and the coefficients and orders of the third order polynomial are coded (39), and then the coded data is obtained. The stored compressed data of the character outline is distributed (40) and decoded by a plurality of decoders (41), and the desired character outline is restored (42) according to the decoding result. The outline of the configuration will be described with reference to FIG. Reference numeral 41 denotes an outline point storage means for storing the position of the outline point Qj of the character image, and 42 denotes a start point Pi and an end point Pi + of an arbitrary outline point Qj among a plurality of outline points Qj constituting the character image contour. By calculating the length of this vector and the amount of deviation between the vector and the contour point with respect to the vector Vi as 1, the vector calculation means for finding a vector Vi whose maximum length li is the maximum when the deviation amount is equal to or less than a predetermined allowable value. (43) is a first comparison means for comparing the length li of the vector Vi with a preset length L, (44) is the intersection angle Qi between the vector Vi and the vector Vi-1 adjacent to the vector Vi; Second comparing means for comparing the preset angle Q. (45) identifies the contour section [Pi, Pi + 1] approximating the vector Vi with a straight line when the comparison result is li > L, and the comparison result is li

When L, the contour section [Pi, Pi + 1] is identified by a curved portion, and the comparison is repeated for the next vector and when li + k> L in the vector Vi + k or Qi + k <Q. Identification means for identifying the contour section [Pi, Pi + k] up to that time as one continuous curved portion, and (46) is a first time starting point Qj of the contour with respect to the contour section [Pi, Pi + k] of the curved portion; Set the sample candidate intervals [Qj, Qj + r], which are sample points of, and arbitrary contour points Qj + r as sample point candidates, and coordinates (xj, yj) of this sample point Qj and sample point candidate Qj + r, curve approximation means for curve approximating the sample candidate interval [Qj, Qj + r] with a cubic polynomial f (x) based on (xj + r, yj + r) and its slopes tj, tj + r, (47) The deviation amount ε calculation means for obtaining the deviation amount ε between the third polynomial f (x) and each lattice point of the sample candidate section. Reference numeral 48 denotes third comparison means for comparing the obtained deviation amount ε and the tolerance Δ, and reference numeral 49 moves the current sample point candidate according to the comparison result and evaluates the updated sample candidate section as described above. The comparison between the quantity ε and the tolerance △ is performed and the comparison result is ε for all contour points.

Satisfying? And setting the sample point candidate when the sample candidate section becomes the maximum as the next sample point, and setting the newly set sample point as the start point of the next sample candidate section, Sample section setting means for sequentially setting the sample section over the entire curved portion and the identified contour section [Pi, Pi + k].

Reference numeral 50 denotes storage means for storing the view coordinates of the linear section and the view coordinates of each sample section set in the curve section, and the coefficients and orders of the third order polynomial approximating each sample section.

Next, the data processing of each part is explained in full detail.

[Image input 30]

By raster scanning on a scanner device or the like, the characters are dot-decomposed onto the x and y matrices, and the bit pattern data obtained therefrom is the character data to be processed.

[Contour extraction (31)]

A dot position (contour point Q) in which two-value data corresponding to the decomposed character data changes from "0" to "1" or "1" to "0" in the x or y direction is obtained. The position of the point is stored in the outline point storage means 41.

FIG. 5 is a diagram showing an example of the character outline obtained in this way, and one character pronounced as "な" (I) is illustrated.

However, when the character image is reproduced by the raster scanning method, the raster image is drawn between the start point and the end point of the scan. Therefore, the information of the contour parallel to the Y axis (raster scanning direction) is not necessarily required.

FIG. 6 omits portions parallel to the Y axis among the contours shown in FIG.

In the figure, the "0" table is a starting point and an end point of each outline, and the narrow space sandwiched between the "0" table is omitted.

As is apparent from the figure, the x coordinate value of each contour is monotonously increasing or decreasing between this starting point and the end point. That is, 1 is a function y whose x is a variable.

Therefore, the contour of FIG. 6 can be obtained by sequentially obtaining the sections (blocks B1, B2, B3 ...) in which the value of the x coordinate monotonously increases or decreases with respect to the contour of FIG.

[Linear approximation (32)]

In the vector calculation means 42 of FIG. 4, a linear approximation is performed on a plurality of vectors Vi which are set such that the length li is maximized within a predetermined tolerance or less.

For example, in FIG. 7, a straight line approximation is performed on a set of vectors Vi (where i = 1 to n-1) of the outline 51 indicated by a dotted line in any one block [P1, Pn]. Connection point coordinate storage unit 60 showing coordinates (xj, yj) of contour point Qj (where j = 1 to m) corresponding to connection point Pi (where i = 1 to n) of each vector Vi in FIG. Remember).

[Identification of straight and curved portions of the contour (33)]

As stated, in general, the outline of a letter has both a straight line and a curved line, and in Japanese Patent Publication No. 58-134745 of the presenter, the straight line part is intended to approximate the entire outline in one block at once. In the portion where the and curve portions are connected, a large number of sample points must be set near the connection point and expressed by a large number of approximation equations, and the amount of data in this portion is increased and the decrease in compression ratio is pointed out.

Thus, the present invention first identifies whether the shape of the contour approximation of the vector approximation is a straight portion or a curved portion according to the length li of the vector determined by the linear approximation described above, and separately processes the straight portion and the curved portion separately. In order to solve the problem, it is possible to obtain an excellent reproduction image even if the character has a complicated contour shape.

FIG. 8 is a block diagram showing one embodiment of the configuration in which the straight portions and the curved portions of the contour are realized, and FIG. 9 is a diagram showing details of main parts of the process in the configuration of FIG.

8 and 9 correspond to the vector calculation means 42, the comparison means 43, 44, and the identification means 45 in FIG.

In Fig. 8, reference numeral 60 denotes a connection point coordinate storage unit for storing the connection points obtained by the linear approximation 32 in units of blocks, and reference numeral 61 sets a linear part identification vector length for presetting the vector length L for identifying. (62) A curve dividing point identification angle setting unit for setting an angle for identifying a point for dividing the curve in advance, and (63) corresponds to the coordinate of the i + 1th connection point Pi + 1, that is, the connection point Pi + 1. The next connection point coordinate register for retaining the coordinates (xj + S, yj + S) of the contour point Qj + S, where 64 is the coordinate (xj, yj) of the contour point Qj corresponding to the connection point Pi of the i The current connection point coordinate register for holding the symbol, 65 denotes the previous connection point coordinate register for holding the coordinates (Xj-u, yj-u) of the curvature point Qj-u corresponding to the i-1th connection point Pi-1, (66) calculates the length li of the vector Vi at each coordinate of the next acceleration point coordinate register 63 and the current connection point coordinate register 64; The vector length calculation unit 67 calculates the angle θ i between the vectors at the connection point Pi at the respective coordinates of the next connection point coordinate register 63, the current connection point coordinate register 64, and the previous connection point coordinate register 65. The angle calculation unit 68 between the vectors is a vector length comparison unit that compares the vector length L set by the straight line identification vector length setting unit 61 with the vector length li calculated by the vector length projection unit 66, ( 69 is an angle comparison unit for comparing the angle θ set by the curve dividing point identification angle setting unit 62 with the angle θ i between the vectors calculated by the angle calculation unit 67 between the vectors, and 70 the vector. It is a linear storage unit for storing the section of the linear portion in accordance with the comparison result of the length comparison unit 68. Reference numeral 71 denotes a curved storage unit for storing the section of the curved portion in accordance with the comparison result of the vector length comparison unit 68 or the comparison result of the vector length comparison unit 68 and the angle comparison unit 69.

Next, the operation will be described. First, angle (theta) is set to the point identification angle setting part 62 which curve-partitions the vector length L in the linear part identification vector length setting part 61, respectively. Next, the connection point coordinate storage unit 60 sends the coordinates of the starting point P1 of any one block [P1, Pn] to the next connection point coordinate register 63. At this point in time, since nothing is stored in the current connection point coordinate register 64, the vector length calculation unit 66, which will be described later, cannot obtain the vector length.

The connection point coordinates stored in the next connection point coordinate register 63 are transferred to the current connection point coordinate register 64, and the coordinates of the next connection point P2 are newly stored in the next acceleration point coordinate register 63. Subsequently, the coordinates of the connection point stored in the current connection point coordinate register 63 are shifted to the current connection point coordinate register 64, respectively, and the coordinates of the next connection point are stored in the connection point coordinate storage unit 60 in the next connection point coordinate register 63. do.

The vector length calculation unit 66 is a coordinate point (xj + S, yj + S) of the i + 1th connection point Pi + 1 of the next connection point coordinate register 63 and the connection point i of the current connection point coordinate register 64. Calculate the length li = (xj + S-xj) ² + (yj + S-yj) ² of the vector Vi from the coordinates of Pi (xj, yj). The calculated length li is compared with the vector length L previously set by the linear part identification vector length setting part 61 by the vector length comparison part 68. FIG.

If li> L, the contour section [Pi, Pi + 1] approximated to the vector Vi is judged as a straight line, and the linear coordinate storage unit 70 uses the start coordinates and the end point coordinates of this section as information about one straight line. Remember to.

L일 때는 윤곽구간 [Pi, Pi+1]을 곡선부로 판단하여 이 구간[Pi, Pi+1]의 시점 좌표 및 종점 좌표를 곡선부에 관한 정보로서 곡선부기억부(71)에 기억한후, 다음의 구간[Pi+1, Pi+2]의 식별을 행한다. 다음의 구간[Pi+1, Pi+2]도 곡선부이면 그때까지 곡선부로 식별해온 구간[Pi, Pi+2]을 연속한 하나의 곡선부로 하고, 그 시점 좌표와 종점 좌표를 하나의 곡선부에 관한 정보로서 곡선 부기억부8719에 기억한다.Again, li

At L, the contour section [Pi, Pi + 1] is judged as a curved section, and the starting point and end point coordinates of this section [Pi, Pi + 1] are stored in the curved storage unit 71 as information on the curved section. Next, the following sections [Pi + 1, Pi + 2] are identified. If the next section [Pi + 1, Pi + 2] is also a curved section, the section [Pi, Pi + 2], which has been identified as a curved section until then, is one continuous curved section, and the starting point and end point coordinates are one curved section. The information stored in the curve additional memory section 8919 is stored in the table.

Then, the above comparison section is repeated until the next contour section is identified as a straight section, and when the vector Vi + k becomes li + k> L, the contour section up to that point already stored in the curve storage unit 71 is already stored. The start coordinates and the end point coordinates corresponding to [Pi, Pi + k] are fixed as information about one curved portion.

As described above, when information on one straight portion or information on the curved portion is obtained, the same identification as described above is repeated for the next vector, and the straight portion forming one block [P1, Pn] is repeated. The data of the and curve portions are stored in the storage portions 70 and 71, respectively. The mode block constituting one character is identified below.

[Segmentation of the outline (34)

On the other hand, in the vector-to-vector angle calculation unit 67, the coordinates (Xj) of the connection points Pi-1, Pi, and Pi + 1 in the difference connection point coordinate register 63, the current connection point coordinate register 64, and the previous connection point coordinate register 65, respectively. -u, yj-u), (xj, yj), (xj + S, yj + S), and the angle between the vectors shown in FIG. 7 (intersection angle of the vector Vi and the adjacent vector Vi-1 θ i is calculated. The calculated angle between the vectors θ i is compared with the angle preset in the angle splitting section identification angle setting section 62 in the angle comparing section 69. And when θ i <θ, the curve division signal is supplied from the angle comparison section 69 to the curve storage unit 71.

In the practice of the present invention, the curve sub-storage unit 71 is operated in accordance with the curve division signal or whether the curve division signal is to be ignored. When a command to ignore the curve division signal is instructed, the curve storage unit 71 faithfully executes the operation described in the section of [Linear portion of the contour and the extinction 33 of the curve portion].

In other cases, i.e., in the state of commanding the operation of operating the curved storage unit 71 according to the curve division signal, for example, the curved sections are identified continuously in the vector Vi, and θ i + 1 <0 in the vector Vi + k. When it is, this curve division signal is supplied to the curve sub storage unit 71. The curve storage unit 71 identifies the connection point Pi + k as a curve dividing point and coordinates the view point corresponding to the contour section [Pi, Pi + k] up to that point already stored in the curve storage unit 71. And end point coordinates are fixed as information about one curved portion.

By the way, unusual points, such as the intersection part of the above-mentioned character image and the tip of a "beep invasion", are usually identified as this curve dividing point. In Japanese Patent Application Laid-Open No. 58-134745, many connection points exist in the vicinity of this unusual point, and it takes time to process the approximation curve obtained, but the contour is divided at the curve division point obtained above, and each curve approximation before and after As a result, the processing time is fast, and an approximate curve can be easily obtained thereon.

FIG. 10 shows an example in which the processes from [linear approximation 32] to [division 34 of outlines] are performed on the outline shown in FIG.

In the drawing, o is a start and end point of one block, o is a connection point obtained by the linear approximation and identification of the curved portion 33, and a curve dividing point.

In addition, the two curve dividing points shown are set because the crossing angle θ n is smaller than the predetermined identification angle.

[Calculation of Slope of Each Contour Point 35]

When approximating the outline shape of the above-mentioned curved part using a cubic polynomial, this cubic polynomial is uniquely determined when the coordinate value and the inclination of two points are determined. Then, first, a method for obtaining the inclination at each contour point on the contour will be described.

In the embodiment of the present invention, the inclination of each contour point is obtained by extracting only a predetermined number of contour points before and after the contour point, and calculating the inclination of the contour point and the slope of the line segment connecting the extracted contour points, respectively. The inclination at the contour point of is calculated in advance and stored.

Next, the calculation method of the inclination at each rim angle is demonstrated according to the example of the curve part shown in FIG.

First, the case where the inclination t1 at the contour point Q1 (starting point of the curved portion) in FIG. 11 (a) is obtained will be explained. Only the arbitrary number of contour points behind the contour point Q1 is extracted, calculating the slope of the line connecting the other contour points, and the slope t1 calculation formula which will be described later in the gradient boundary point Q ₁ in each of these segments.

For example, in the case of obtaining the inclination by extracting two contour points on the rear side of the contour point Q1 to be inclined, first, the inclination m12 of the line part Q1 Q2 and the inclination m13 of the line segment Q1 Q3 are obtained.

The slope m12 of the line segment is at two coordinates (x1, y1), (x2, y2)

You can get it by In addition, the case of the inclination m13 is calculated | required similarly.

When the slopes m12 and m13 of the line segments are obtained, the slope t1 at the contour point Q1 is

Obtain by

Next, the case where the inclination t2 at the next contour point Q2 is calculated | required according to FIG. 11 (b) is demonstrated.

In this case, only the predetermined number of contour points which exist before and after contour point Q2 are extracted, and the inclination of the line segment which connected each contour point in contour point Q2 is respectively calculated | required, and it is set as the inclination t2 in contour point Q2.

For example, when the contour point is extracted two points front and back to obtain the inclination, only one point of Q1 can be extracted from the front side at the contour point Q. In this case, the slope t2 is obtained by the following equation by extracting one point front and back.

If the specified number of contour points does not exist as described above, the slope is obtained by extracting the maximum number of contour points within the specified range.

Next, the case where the inclination t3 at contour point Q3 is calculated | required like FIG. 11 (c) is demonstrated. In this case, in the contour point Q3, the predetermined number of contour points exist before and after. Therefore, the slopes m13, m23, m34, and m35 of each line segment connecting the contour point Q3 and the other contour points are calculated to calculate the slope t3 at the contour point Q3,

Obtain from

Then, in the same manner, the inclination at each contour point is calculated, and the inclinations tn and tn-1 at the end point Qn of the curved portion or the contour point Qn-1 adjacent thereto are shown in Figs. 11 (a) and 11 ( It calculates like the case where the inclination of the viewpoint Q1 shown in b) or the contour point Q2 adjacent to it is calculated | required.

In this way, the calculation of the inclination at each contour point is performed by extracting only a predetermined number of contour points related to the front and rear of the contour point. The above description is an example when the predetermined number = 2.

[Determination of Sample Point (36)]

When the inclination at each contour point which comprises a curved part is calculated | required as mentioned above, a sample section is set on the outline corresponding to a curved part next.

The sample section is set by the curve approximation means 46, the deviation amount calculation means 47, the comparison means 48, and the sample section setting means 49 in FIG.

12 is a view showing the process.

Therefore, first, two contour points Qj and Qj + r are sequentially selected in order to set the sample interval (see FIG. 13).

In the following description, among these two contour points, the front contour point Qj is called a sample point, and the rear contour point Qj + r is called a sample point candidate. The section using these two points as the starting point and the ending point is called a sample candidate section [Qj, Qj + r].

The sample point Qj is set at the start of the curve portion as an initial condition, and the sample point candidate Qj + r is set at a position separated by r contour points from the sample point Qj.

So first, based on the coordinates (xj, yj), (xj + r, yj + r) of these sample points Qj and sample point candidate Qj + r, and their slopes tj, tj + r, Coefficients and orders of the third order polynomial f (x) are obtained, and each deviation amount ε between all contour points existing in the sample candidate interval [Qj, Qj + r] and the approximation curve obtained by the third order polynomial is predetermined. By comparing whether or not the tolerance △ or less, the validity of the obtained approximation curve is determined.

△의 경우, 상기 구해진 근사곡선을 허용할수 있는 것으로 판단한다.Now, for all contour points of the sample candidate section [Qj, Qj + r], ε

In the case of Δ, it is determined that the obtained approximation curve is acceptable.

By the way, the longer the interval that can be expressed by one approximation curve, in other words, the smaller the number of approximation curves required to express a given arbitrary contour, the smaller the number of data and the more efficiently it can be compressed.

△일때, 현재의 샘플점후보 Qj+r에 인접하는 윤곽점 Qj+r+1을 새로운 샘플점후보로하여 상기 비교를 하고, 이후 비교 결과가 하나라도 ε＞△로 되기까지 샘플점 후보를 순차 갱신하여 상기 비교를 반복하고, 윤곽점 Qj+r+p에서 하나라도 ε＞△로 되었을때, 그 직전의 윤곽점 Qj+r+p-1을 제2의 샘플점으로하여 샘플구간 [Qj, Qj+r+p-1]을 설정하고, 이 샘플구간 [Qj, Qj+r+p-1]에 있어서의 근사 곡선이 주어진임의의 윤곽을 표현하는데 필요한 근사곡선의 하나로서 결정된다.Thus, the approximation curves obtained as described above with respect to the sample candidate sections [Qj, Qj + r], which are determined to be the sample point candidates of the arbitrary contour point Qj + r initially set, and the sample candidate sections [Qj, Qj +]. r] is calculated and the result of comparison with the tolerance △ is calculated.

When [triangle | delta] is made, the said comparison is made using the contour point Qj + r + 1 adjacent to the current sample point candidate Qj + r as a new sample point candidate, and the sample point candidates are sequentially made until any comparison result becomes [epsilon]> △. When the above comparison is repeated and at least one of the contour points Qj + r + p becomes ε> Δ, the immediately preceding contour point Qj + r + p-1 is used as the second sample point. Qj + r + p-1] is set, and an approximation curve in this sample section [Qj, Qj + r + p-1] is determined as one of the approximation curves required to express a given arbitrary outline.

△ 되기까지 샘플점후보를 순차 갱신하여 상기 비교를 반복하여, 윤곽점 Qj+r-p에서 모두 ε

△로 되었을때, 이 샘플점후보 Qj+r-p를 제2의 샘플점으로 하여 샘플구간 [Qj, Qj+r-p]을 설정하고, 이 샘플구간 [Qj, Qj+r-p]에 있어서의 근사곡선이, 주어진 임의의 윤곽을 표현하는데 필요한 근사곡선의 하나로서 결정된다.On the other hand, the approximation curves obtained as described above with respect to the sample candidate sections [Qj, Qj + r], which are determined as sample point candidates, which are initially set to arbitrary contour points Qj + r, and these sample candidate sections [Qj, Qj]. The deviation amount ε with each contour point of + r] is calculated, and when ε> △ with only one result compared with the tolerance △, the contour point Qj + r-1 adjacent to the current sample point candidate Qj + r is calculated. The comparison is made as a new sample point candidate, and the comparison results are then ε

The sample point candidates are sequentially updated until △, and the comparison is repeated, and both at the contour point Qj + rp

When [triangle | delta] is set, the sample section [Qj, Qj + rp] is set using this sample point candidate Qj + rp as a 2nd sample point, and the approximation curve in this sample section [Qj, Qj + rp] becomes, It is determined as one of the approximation curves needed to represent a given contour.

By such a process, the contour is divided in the sample section while determining the longest sample section within the range of the deviation?. The outline is approximated by a cubic polynomial determined by this sample interval.

Next, the calculation of the approximation curve and the calculation method of the amount of deviation are explained based on FIG.

In FIG. 13 (a), reference numeral 90 denotes a contour, B denotes a contour point on the contour 90, 91 denotes an approximation curve of the cubic polynomial f (x), and 92 denotes a sample candidate interval [Qj, Qj + r] is a straight line connecting two points between the starting point Qj and the end point Qj + r, (93) is the vertical line lowered from the point B to the straight line 92, C is the intersection of the approximation curve 91 and the vertical line 93, BA is the amount of deflection ε x in the x direction between the contour 90 at the point B and the approximation curve 91, and BC is the amount of deflection in the y direction between the contour 90 and the approximation curve 91. ε y.

The slopes tj, tj + r and the coordinate values (xj, yj) (xj + r, yj + r) of the already calculated time point Qj and sample point candidate Qj + r are substituted into the following equation.

bj = tj

When the approximation curve is obtained as described above, the amount of deviation ε between the approximation curve and each contour point is obtained, respectively. FIG. 13 (b) enlarges the amount of deflection epsilon between the outline 90 and the approximation curve 91 in FIG. 13 (a). In the figure, the line segment CA is assumed to be parallel to the straight line 92, and the inclination of the straight line 92 is the inclination m of the line segment CA.

And as shown, when 1 ml> 1, when the deviation amount (epsilon) X with respect to the X direction of the point B and the approximation curve 91 is calculated | required, the desired deviation amount (epsilon) will be

Can be obtained as

1일 때는 마찬가지로And although not shown in the drawing 1ml

When it is 1

The amount of deviation desired can be found at.

In the above description, if the amount of deviation ε in the next sample candidate section [Qj, Qj + r + p] exceeds the tolerance, the previous sample point candidate Qj + r + p-1 is used as the sample point. Sample intervals were determined.

However, as another example, when the tolerance is exceeded, the sample point candidate Qj + r + p is memorously stored at that point, and the evaluation of the deviation amount ε when the sample candidate section is made up to several points before Do it. When the condition of the tolerance is satisfied in the sample candidate section in which the sample point candidate is advanced several points, the candidate section is corrected to a new sample section, and the above-described evaluation process is repeatedly executed while advancing the sample point. By performing such an earlier writing, the sample interval can be made longer, and the data compression rate is improved.

Of course, when the tolerance is reached even before the predetermined number of points, the sample interval is determined using Qj + r + p-1 as the sample point.

[Sloped Reproduction (37) of Each Contour Point]

Since the inclination of each contour point obtained by the [inclination calculation 35 of each contour point] is obtained by quantizing the contour point for obtaining the inclination and the predetermined number of contour points before and after the contour point from the inclination of each line segment, It may be different from the inclination of the actual outline. Therefore, in order to approximate the outline more faithfully as necessary, the slope is calculated again as follows. That is, the inclination obtained by the first derivative f '(x) of the tertiary polynomial f (x) in each sample section is the inclination at each contour point of this sample section.

In the embodiment of the present invention, after each sample section is determined, all slopes at the contour points of each sample section are obtained and stored in advance.

[Recrystallization of Sample Point (38)]

When the inclination at each contour point is calculated again by the [inclination recalculation 37 of each contour point], the said [determination 36 of a sample point] is used using the inclination in each calculated contour point. ], The sample section determined again while determining the sample point is approximated by a third-order polynomial f (x) that is more faithful to the contour. In addition, since the number of sample intervals can be reduced, the data compression rate is improved.

By processing the above [inclined recalculation 37 of each contour point] and [recrystallization of the sample point 38], the contour can be approximated more faithfully, and the data compression ratio is also improved. It takes time.

If the character is a simple character, the character outline is faithfully approximated, and thus the above processing is not necessary.

FIG. 14 shows an example in which the processes from [decision 36 of sample point] to [re-determination 38 of sample point] are performed on the outline shown in FIG.

는 1블록의 시점 및 종점, ▲는 접속점, ●는 곡선 분할점,

는 직선부, 무표시는 곡선부를 나타내고 있으며, △가 새로 설정된 샘플점이다.Like in Figure 10,

Is the start and end point of 1 block, ▲ is the connection point, ● is the curve division point,

Denotes a straight portion and no mark denotes a curved portion, and Δ is a newly set sample point.

[Encoding 39] and [Storage of Compressed Data 40]

Each linear part stored in the linear part storage unit 70 of FIG. 8 is regarded as a sample section, and the viewpoint coordinates are encoded.

Subsequently, the viewpoint coordinates of the respective sample sections in the curve section obtained by the [determination 36 of the sample point] or [the recrystallization 38 of the sample point], and the angle of the approximation curve represented by the cubic polynomial f (x) Coefficients and orders are coded separately.

By storing these coded data in the storage means 50 of FIG. 4, compressed data faithful to the contour can be obtained.

FIG. 15 is a diagram showing an example of a preferable format of any one block of data to be applied in the practice of the present invention as described above.

In the format of the drawing, the block header stores the coordinates of the end point of one block and the number of sample sections of the straight and curved portions present in the one block, and the segment header contains the coordinates of the start point of the one sample section, and the third order polynomial f The order of x) is stored, and the segment information stores each coefficient of the third order polynomial f (x) obtained by the above order.

The coded data of the sample section constituted by the segment header and the segment information is arrayed by the number of sample sections stored in the block header to form block data for specifying the outline shape of one block.

The data is structured in the same format for the other blocks, and is compressed data specifying the outline shape of one molecule as a whole.

Next, when storing the block data coded by the above format, the decoding process time for each block is in the order of the longest block, in the order of long time, for example, in the case of speedy restoration of the contour. The storage method for storing block data in the storage means 50 will be described.

For example, FIG. 16 (a) shows each block of the letter "あ" (in the description, each block is numbered 1 to 20).

The memory is stored in order from the long time required for decryption (for example, the distance between the start point and the end point of each block in the X direction). In the example of FIG. 16 (a), the longest time required for decoding one block is the block 12, which is followed by the block 10 and the block 16, and the block 3 is the shortest.

Therefore, the compressed data for one character finally obtained by this example is stored as a set of block data as shown in FIG. 16 (b).

[Distributed Decoder 41]

What has been described so far is the creation of compressed data specifying contours.

Next, a method of restoring the original outline will be described.

In general, restoration of an outline is performed as follows. First, block data is transmitted to the decoder in block data units. The decoder then restores this block data. When the restoration of one block data is completed, block data that has not been restored is sequentially transmitted to the decoder. In this way, the decoder decompresses the block data for one character to restore the compressed data to the outline.

Next, as shown in FIG. 16 (b), the case of the compressed data stored in the order of the long time required for decoding the block data for one character is explained. In this case, an embodiment in which the block data for each block is sequentially transmitted to a plurality of decoders for processing in order to restore the contour, and the next block data is processed by the processed decoder will be described.

Next, a description will be given based on FIG. 17 and FIG. 18. Fig. 17 is a block diagram showing one embodiment for contour restoration in the present invention. In the figure, reference numeral 120 denotes a compressed data storage unit for storing each block data in order of long time required for decoding one block, storage means 50 in FIG. 4, and 121 denotes one block data described later. A selector 1 for selecting and transmitting a decoder, (122) a magnification storage unit for storing a desired input magnification separately, (123) is a decoder group consisting of n decoders for restoring the block data to a contour corresponding to the magnification, (124) Selector 2, 125 which selects a decoder whose processing has been completed from the decoder group 123 and transmits the decoded outline data to the one-character storage unit 125 described later. The one-character memory unit 126 for storing the sent outline data is an output device for printing or displaying characters on the basis of the one-character outline data stored in the one-character memory unit 125.

Next, the operation will be described.

In the compressed data storage unit 120, the time required for decoding one block is transmitted to the selector 1 121 in a long order (for example, in the order of blocks 12, 10, block 16, ... in FIG. 16). . The selector 1 121 selects and transmits the transmitted block data in order from the decoder 1 of the decoder group 123. The decoder which has finished the transmission of the block data starts the reconstruction processing of the outline based on the block data and the desired magnification data stored separately in the magnification storage section 122. The selector 2 124 sequentially selects the decoders for which the restoration process has been completed, and transmits the deconstructed contour data to the one-character storage unit 125. The decoder which has completed one block restoration process requests the selector 1 121 to transmit block data, and the next block data for one block is transmitted.

FIG. 18 is a timing chart showing an example of the processing state of each decoder group 123 shown in FIG.

In the figure, T1 denotes a time for transmitting block data from the selector 1 121 to the liver decoder, and T2 transmits contour data decoded by each decoder to the one-character memory 125 through the selector 2 124. Is the time, and T3 is the decoding process time of block data in each decoder. T4 represents the processing time for one character.

When n blocks of block data are transferred to the decoders 1 to n and processing starts, the n + 1th block data is then transferred to the decoder having the shortest processing time (in this case, decoder 3) and processing continues. .

In addition, since each block data to be transmitted is arranged in order of decreasing processing time as described above, as shown in FIG. 18, each decoder ends processing at the same time and the decoding for one character is completed. . That is, since each decoder operates evenly and proceeds to decode the block data, the decoder starts decoding of the next character block data without waiting for another decoder for a long time until the decoding operation of one decoder is completed. By distributing to the decoder and supplying n blocks of data, the entire decoding time can be shortened.

[Reconstruction of the outline (42)]

As described above, the contour data of one block restored by each decoder stores the contour data of one character stored in the one-character storage unit 125, for example, a laser beam printer, a CRT photographic magnet or various display devices. The desired characters are restored by supplying them to the output device 126 of.

As described above, according to the present invention, the contour is identified by the straight portion and the curved portion, the straight portion is represented by the straight line, and the curved portion is represented by the curve approximation, so that the character image can be faithfully reproduced. In addition, since the curve portion is sequentially divided in the sample section that can be most appropriately curved, the contour can be faithfully reproduced with a small sample score. In addition, the nth-order polynomial that approximates each sample section is calculated for each sample section considering the inclination of the start point and the end point of each sample section, and thus can be processed at high speed without complicated calculation.

In addition, since the outline is approximated by a cubic polynomial, the deterioration of the character image quality at the time of high magnification conversion, which is a drawback of the conventional vector approximation method, can be eliminated.

As a result of applying the character image data processing system according to the present invention described above to the famous Hiragana character "な" composed of 800x800 dots, the data compression ratio of 1.21% is obtained when the tolerance for the desired character image is 1 dot. Could get

In addition, although the above description is a description in the case of protecting and compressing characters, it is clear that various marks, symbols, pictures, and the like other than the characters can be handled in the same manner.

According to the present invention, it is possible to provide a character image data processing system capable of obtaining data at a high speed and sufficiently storing contour smoothness at a sufficiently high compression rule.

Claims (2)

In a character image data processing system configured to store coded data for specifying a character image contour, a vector Vj in which one of a plurality of contour points Qj constituting the character image and the contour is defined as a start point Pj and an end point Pj + 1. A vector calculating means 42 for obtaining the length of this vector and the amount of deviation between the vector and the contour point so as to obtain the vector Vj whose deviation is equal to or less than a predetermined allowable value and whose length lj is maximum; First comparing means (43) for comparing the length lj of the vector Vj with a preset length L; Second comparing means (44) for comparing the intersection angle θ j between the vector Vj and the vector Vj-1 adjacent to the vector Vj and the preset angle θ; When the comparison result is lj> L, the contour section [Pj, Pj + 1] approximating the vector Vj is identified by a straight line portion, and the comparison result is lj.

When L, the contour section [Pj, Pj + 1] is identified by a curved portion, and θ j

When θ, the comparison is repeated for the next vector, and when the vector Vj + k becomes lj + k> L, the contour section [Pj, Pj + k] up to that point is identified as one continuous curve portion. Identification means 45; The contour section [Pj, Pj + k] of the curved portion is a sample candidate section [Qj, Qj + r] having the starting point Qj of the contour as the first sample point and an arbitrary contour point Qj + r as the sample point candidate. And based on the coordinates (Xj, yj) (Xj + r, yj + r) of this sample point Qj and sample point candidate Qj + r, and their slopes tj, tj + r, the sample candidate interval [Qj, Qj + r] to obtain a third-order polynomial f (x) approximating the curve, and then calculate the deviation ε between the third-order polynomial f (x) and each contour point of the sample candidate section, and calculate the deviation ε and the allowance Third comparing means 48 for comparing the error?; And a storage means (50) for storing the viewpoint coordinates of the linear section, the viewpoint coordinates of each sample section set in the curve section, and coefficients and orders of the third order polynomial approximating each sample section. Processing system. The third comparison means (48) according to claim 1, wherein the third comparison means (48) configured to obtain a cubic polynomial f (x) that approximates the sample candidate interval [Qj, Qj + r] is first used as the contour section [Pj, Pj] + k] on the basis of the average value of the inclination of each line segment connecting a predetermined number of contour points located before and after the contour point, the slope of the contour point is calculated, and then the slope of the calculated contour point and Based on the coordinate values of the contour points, a plurality of third order polynomials f (x) approximating the contour section for each part are obtained, and again based on the first derivative f '(x) of each third order polynomial f (x). The third order polynomial f, which obtains the inclination t of all contour points in the contour section yarn in advance and curves the sample candidate section [Qj, Qj + r] using the slope t of the previously obtained contour point, x) to obtain a text image data processing system.

Images