JPS6219975A - Graphic form processing system - Google Patents

Abstract

PURPOSE:To obtain image data related to characters from a handwritten graphic form and to make information compression of an original graphic form by making a handwritten graphic form into fine lines and separating characters, broken lines and solid lines basing on length between local feature points such as intersections and endpoints etc. CONSTITUTION:A fine line making section 1 makes inputted graphic from into fine lines using a fine line making mask, and a local feature extracting section 2 extracts local intersections, endpoints etc. from the fine line graphic form using an intersection extracting mask. A stroke following section 3 makes stroke pursuit starting from local intersections, and extracts local inflection points and local endpoints and further, determines length of each line element. A global feature extracting section 4 compares length of line element and threshold value, and disposes of unnecessary local inflection points and local endpoints, and disposes of local intersections which are too close to each other, and determines global intersections, global inflection points and global endpoints and performs information compression. A character separating section 5 judges local feature points of short distance as parts of characters and those of long distance as parts of graphic form and separates characters and graphic forms.

Description

[Detailed Description of the Invention] [Summary] In a graphic processing system that compresses and expresses a figure drawn with a mixture of characters, broken lines, solid lines, etc. based on local feature points, starting from the intersection of thin line figures, The distance between the local inflection point and the line element is extracted by tracing the stroke, and the global inflection point is determined from the obtained local inflection point to the extent that the characteristics of the drawn figure are not impaired. In addition to compressing information, character information is separated based on the sum of distances for each stroke.

[Conventional technology]

The conventional graphic processing system is 1, for example, 1982, 295.
As stated in specification No. 45.

The figure to be compressed is thinned and approximated by a broken line, regardless of the character or the like, and the information is compressed to the extent that the characteristics of the original figure are not lost.

[Problem that the invention seeks to solve]

In the conventional graphic processing system, information was compressed by thinning the figure and using broken line approximation regardless of whether it was two figures or one character, so there was a problem that image data corresponding to a character could not be extracted from among five figures.

[Means for solving problems]

In order to solve the above problem, the present invention calculates the distance between local feature points extracted by thinning the original figure, and if this distance is shorter than a preset threshold, it is a character part,
By adopting a configuration that separates letters and figures by determining that long objects are graphic parts, information is compressed for two figure parts to the extent that the characteristics of the original figure are not lost, and information is compressed for one character part. In this case, character data can be separated from the original figure and processed.

FIG. 1 shows the basic configuration of the present invention. In the figure, 1 represents a thinning processing section, 2 a local feature point extraction section, 3 a stroke tracking processing section, 4 a global feature extraction processing section, and 5 a character separation processing section.

In FIG. 1, a thinning processing unit 1 thins an input figure by applying a well-known thinning mask as described in Japanese Patent Publication No. 29545 of 1981. It is something. For example, a thin line figure as shown in FIG. 2 is obtained.

The local feature extraction unit 2 extracts local intersection points, end points, etc. from a thin line figure using an intersection extraction mask.

The stroke tracing processing section 3 performs stroke tracing starting from local intersection points, extracts local inflection points and local end points related to the stroke, and determines the length of each line element.

The global feature extraction processing unit 4 firstly compares the length of the line element with a certain threshold value, sorts out unnecessary ones among the local inflection points and local end points, and secondly, Thirdly, by sorting out intersection points that are too close together, we can find global intersection points and global inflection points.

The fourth step is to determine global endpoints and perform information compression.

The character separation processing unit 5 separates characters from the original figure, and determines that a short distance between local feature points is a character part, and a long distance is determined to be a figure part, and distinguishes it from a character. It is designed to separate the figure from the figure.

[Effect]

A human figure is made into thin lines, stroke tracing is performed starting from the intersection of the thinned figures, and the distance between the three local inflection points and the line elements is extracted, and those whose distance is shorter than a preset threshold are used as text parts. It is determined that the long part is the figure part.

Then, for the part judged to be the 2-figure part.

Based on the local feature points, the information is compressed and expressed to the extent that the features of the original figure are not distorted. Furthermore, a portion determined to be a two-character portion can be expressed using one-character information or the like.

〔Example〕

FIG. 2 is an explanatory diagram of a graphic example explaining stroke tracking processing, FIG. 3 is an explanatory diagram of a table showing local inflection points and distances, and FIG. 4 is an explanatory diagram of a table explaining feature points. .

The thin line figure shown in FIG.
is generated from the input figure by applying a well-known thinning mask (for example, Japanese Patent Publication No. 29545 of 1982). The "◎" marks in the figure indicate local intersection points, and the "■" marks in one figure indicate local inflection points. 1t...l,, indicates the distance between local inflection points. In addition, local intersection points indicated by A and B in FIG. 1 indicate one stroke, which is a basic unit of broken line approximation.

In FIG. 2, for example, a thin line figure is traced from a local intersection point in the figure to a local intersection point B, and a table shown in FIG. 3 is created. Next, determine the global inflection point using the coordinates (X,, Y,), (Xz, Yz) of the local inflection points shown in this table and the distances between the local inflection points. . This determination is executed by the global feature extraction processing unit 4 as described above.

Next, the following processing is executed to extract character information.

(1) Find the sum of the distances between local inflection points.

(2) Compare this total L with a preset threshold Th.

(3) If ≦T6, set a character flag indicating that it is a character part for the 2 local feature points, a pointer between local feature points, and a storage table pointer that stores the final approximate vector. (Figure 4).

FIG. 4 is a table explaining feature points. ■ in the figure indicates the coordinates of local feature points. ■ in the figure shows.

The direction of the stroke emanating from the position specified by the coordinate values stored in ``■'' in the figure is stored.

For example, for the coordinate value (0,0), that is, the local feature point (intersection point) indicated by "A" in FIG. 2, R (rightward),
"1" is set in each of U (upward direction) and D (downward direction).

■ in the figure indicates a character flag, and if the strokes coming out from one local feature point in each direction are shorter than a predetermined value, "1" is displayed.
” is cented. For example, if the stroke extending to the right from the local feature point is shorter than a predetermined value, the one-character flag FR is set to “1”.

In the figure, ``■'' indicates a pointer, and ``■NO'' in the illustration of another local minutiae connected to the own local minutiae is set.

■ in the figure indicates a pointer to the storage table, Sm, S
u and St-So indicate the starting point of the line segment corresponding to the final approximation result of the stroke coming out from the local feature point, and E, Eu, Et,
Eo indicates the end point.

Next, the procedure for separating one character will be explained.

fl+ Character flags F*, Fu for each local feature point.

Search FL, F, and if the strokes coming from the two local feature points are all "1", set the 1 character flag F to "1"
Set to .

(2) If the pointer PR, PLl+ PL, PD of a local minutiae point whose character flag Fr is "1" is traced and a local minutiae point whose one character flag F is not "1" is reached, the character in this process All local feature points whose flag F1 is "1" are set to "0".

Otherwise, the 7-character flag FT is left as "1".

A local feature point corresponding to the extracted character part is set as a character flag F, "■", from among the thin line figures thinned by the above processing. Next, this character flag F
7 Using the ■ storage table pointer in the diagram corresponding to the local feature point set as "1",
51944), which is not shown, is accessed to separate the broken line from one character. At this time, the character flag FT of the end point (local feature point) of the minute line element constituting the broken line is set to "2" to distinguish it from the character flag 1 of the one character portion.

Through the above processing, characters and figures (solid lines, solid lines,
(dashed line), and each is approximated by a broken line.

Furthermore, regarding a single character part, image data of a single character can be extracted based on the coordinates of local feature points connected to the single character flag FT.

5 to 8 show configuration diagrams of one embodiment of the present invention.

FIG. 5 shows an embodiment of the configuration of the thinning processing section 1 used in the present invention. Numerals 18 and 19 in the figure are sub-
20 is an address control unit that generates an access address for input video 21 according to instructions from sub-channel 18; 21 is an input video corresponding to image information; 22 is a module control unit; The program module operates a logic filter circuit (thinning mask) 23, which will be described later, in response to an instruction from the sub-channel 19.

23 is a logic filter circuit (thinning mask) that performs thinning based on the video data sequentially read out as the input video 21; 24 is a thin line figure that is thinned by the logic filter circuit 23; The thin line figures shown in FIG.

FIG. 6 shows the configuration of one embodiment of the local feature point extracting section 2. As shown in FIG. In the figure, 18-1 and 19-1 are sub-channels, and 20-1 is an address control unit.

24 is a thin line figure corresponding to FIG. 5; 25 is a module control unit that operates a program module that puts a logic filter circuit 26 (to be described later) into operation according to instructions from the sub chaffle 19-1; 26 is a logic filter circuit which extracts local feature points based on the thin line figure data sequentially read out as the thin line figure 24; 27 is a feature point video which extracts the extracted local feature points as described above. It represents what is held corresponding to the address where the thin line figure 24 exists.

FIG. 7 shows an embodiment of a local feature point coordinate extraction processing section, which extracts the coordinate position of each local feature point based on the feature point video 27 obtained by the configuration shown in FIG. . The numbers 18-2 and 19-2 in the figure are sub channels, respectively, and 20-2 is an address system? The feature point video 228 corresponding to FIG. 29 is a coordinate point extraction circuit which, when a local feature point in the feature point bidet No. 27 is read out, extracts the local feature point and obtains the coordinate information of the feature point. This shows what is set in the feature point table (■, ■) in Figure 4.

FIG. 8 shows the configuration of an embodiment of the information compression processing section and the character separation processing section 5. In FIG. Reference numerals 1.8-3 and 19-3 in the figure indicate sub-channels, 20-3 an address control section, and 24 a thin line figure corresponding to FIG.

30 is a feature point table corresponding to FIG. 7, 31 is a microprocessor, and 32 is a vector storage table for storing figures to be processed, coordinate points, etc.

A thin line figure 24 is generated from the one-man-powered bidet 21 according to the configuration shown in FIGS. 5 to 8 described above, and then local feature points are extracted from this thin line figure 24. And 2
A feature point table 30 (shown in FIG. 4) is generated from the local feature points.

The microprocessor 31, which has sequentially read out the contents from the feature point table 30, executes the processing explained using FIGS. Points, end points) and mutual connection states are stored in the vector storage table 32, and information for extracting one character as image data is set in the vector storage table 32.

〔Effect of the invention〕

As explained above, according to one aspect of the present invention, a free handwritten figure drawn in m, in which characters, broken lines, solid lines, etc. are mixed, is thinned, and the intersections in the thinned figure are thinned.

Since we have adopted a configuration in which end points, etc. are recognized as local feature points and two characters, broken lines, and solid lines are separated based on the length between these local feature points, images related to characters separated from free handwritten figures are used. In addition to being able to obtain data, information on the original figure can be compressed to the extent that the characteristics of the original figure are not lost.

[Brief explanation of the drawing]

FIG. 1 is a diagram illustrating the basic configuration of the present invention, and FIG. 2 is an explanatory diagram of a graphical example for explaining stroke tracking processing. FIG. 3 is an explanatory diagram of a table showing local inflection points and distances. FIG. 4 is an explanatory diagram of a table for explaining feature points. FIGS. 5 through 8 show configuration diagrams of one embodiment of the present invention. In the figure, 1 is a thinning processing section, 2 is a local feature point extraction section, and 3 is a thinning processing section.
4 is a stroke tracking processing unit, 4 is a global feature extraction processing unit,
5 represents a character separation processing section.

Claims (1)

[Claims] A graphic processing system that expresses a drawn graphic by compressing graphic information based on local feature points, comprising: a thinning processing unit (1) that thins a graphic; Local feature point extraction unit (2) that extracts local feature points from the thin line figure obtained by (1)
Then, at least one stroke of the thinned figure is traced from the local feature point obtained by this local feature point extraction unit (2), and the local feature point on the stroke is traced until reaching another local feature point. a stroke tracing processing unit (3) that extracts inflection points and the distance of line elements between the local inflection points or between the local feature points and the local inflection points; a global feature extraction processing unit (4) that compares the distance with a threshold and extracts all adjacent local inflection points within this threshold as a global inflection point; It is equipped with a character separation processing unit (5) that calculates a sum and compares this sum with a threshold value to separate character parts, and calculates strokes between local feature points using feature point coordinates and global inflection point coordinates. 1. A graphic processing system characterized in that the graphic processing system is configured such that character data can be expressed and character parts can be extracted using the coordinates of a circumscribed rectangle of character data.