JPS63172387A - System for separating and reading character and symbol - Google Patents

Abstract

PURPOSE:To easily separate characters, symbols, and graphic forms by processing color parts of a reduced image which is generated by selecting an input image having a smaller binarization threshold as entirely black parts. CONSTITUTION:A user paints out a character part in a drawing with a trans lucent color pen, etc., as shown in a figure. This color pen uses such a color that when a scanner 1 makes a read, the variable density level is between THlow and THhigh. The drawing whose character part is colored as mentioned above is inputted to the scanner 1 and then an image which is converted into a binarized signal so that black lines are black, the white background is white, and the color part is checkered. A reduction part 3 reduces the image contents of the buffer 2 to a half vertically and horizontally. This reduction is carried out by two kinds of methods, and while one system A reduces the checkered pattern part in intermediate color into the entire black part, the other system B reduce only the character constitution part of the color part into black picture elements.

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention relates to one drawing recognition process, and in particular to a character/symbol separation reading method that separates and reads graphic parts and characters/symbol parts.

C. Prior Art] Conventionally, in this type of separation reading system for the letter e symbol, in drawing recognition, it is necessary to separate and identify, for example, dimension 1, description 5, etc. in the drawing from the entire drawing.

In order to separate these characters and symbols from other graphical elements as shown in FIGS. 18(A) and (B), all line elements in the image must be analyzed to determine the length of continuous line segments, A method was needed to extract data such as branching states of line segments and make decisions.

The operation of the conventional method will be explained based on FIGS. 18(A) and 18(B). First, in Figures 16(A) and 16(B), the input image side (shown in Figure 16(A)) contains a mixture of graphic elements and text elements, so these are used for recognition. To separate the elements (as shown in Figure 16(B)), for example, by tracking consecutive black pixels, labeling each unique consecutive element and counting the number of black dots in each element. If the number is larger than a certain threshold, it is determined that it is a figure, and if it is smaller than a certain threshold, it is determined that it is a letter, thereby identifying figures and characters.

[Problems to be Solved by the Invention] Conventional character/symbol separation recognition methods are based on the size of individual continuous elements in an image as described above, so if there is a broken line in the image, for example, .

When there are small geometric elements, it is impossible to identify them based on size alone, and it is necessary to further analyze the element's other characteristics and relationships with surrounding elements, which requires a very complex algorithm. It had the problem of becoming

This invention was developed to solve the above-mentioned problems, and the purpose is to obtain a processing method for separating and recognizing specific characters and symbols in a drawing by using a reliable and simple algorithm. shall be.

[Means for Solving the Problems] The method for separating characters and symbols in drawings according to the present invention is that after an operator performs a simple process of marking specific characters and symbols in a drawing with a colored pen cap, , a scanner with a dither pattern in which two binarization thresholds, large and small, are arranged alternately vertically and horizontally for each pixel, is used to scan the image, and the larger and smaller binarization thresholds are selected to create two types of reduced images. The colored parts are identified by

[Operation] In the present invention, the colored part is fully exposed in the reduced image created by selecting the smaller binarization threshold from the input image,
Reduced images created by selecting the larger binarization threshold are not affected by coloring, so colored text/symbol areas are detected from the former reduced image, and characters/symbols are detected by comparing them with the latter reduced image. and separate the shapes.

[Embodiment] An embodiment of the present invention will be described below with reference to FIGS. 1 to 13. FIG. 1 is a block diagram of an embodiment of the present invention, and FIG. 2 shows the scanner (1) in FIG.
) dither pattern, Figure 3 (A), (B),
(C) White obtained with the dither pattern shown in Fig. 2 above;
Binarized image pattern of black and intermediate colors, Figure 4 is a table of figures to be input, Figure 5 is a table of figures after coloring, Figure 6 is an image pattern of the colored part, Figure 7 is a reduction of the image. pattern, 8th
Tables and diagrams based on reduction methods (A) and (B) are shown in FIG.
Figure 1θ is a reduced table of intermediate color patterns, Figure 11 (A)
, (B) show the figures after reduction, and FIGS. 12 and 13 show the figures after the reduction.

In Figure 1 above, (1) is a scanner that binarizes the figure to be processed into black as "l" and white as "0" by photoelectric conversion, and the binarization threshold can be set by a dither pattern, and (2) is a scanner that A buffer that stores the image read in the above scan (1), (3) a reduction unit that reduces the image in the buffer (2) by percentage in two ways according to a set dither pattern, (4)
] is a buffer that stores the image that has been reduced in two ways in the reduction unit (3), the one in which the character/symbol area becomes clearer;
(5) is a buffer that stores the other station line small image other than the reduced image stored in buffer (4), and (6) processes the image in buffer (4) to detect the character φ symbol area. (7) is a character/symbol recognition unit that performs recognition processing on the character/symbol area detected by the detection unit (6) from the image in the buffer (5), (8) is the buffer (5) (8) is a synthesis unit that combines the processing results of the character/symbol recognition unit (7) and the graphic processing unit (8). The processing section (10) is a main control section that controls each of the above sections.

Figure 2 shows the dither pattern of the binarization threshold set in the scanner (1), with a large binarization threshold (THhiqh) "6" and a small binarization threshold (THhiqh) "l". They are arranged alternately vertically and horizontally at the middle of one pixel. T
The values of Hh + qh = 6 and THIow = 1 are an example where white is level 0 and black is level 7 when gradation is photoelectrically converted.Actually, the values should be set at appropriate intervals between white and black gradations. good.

In this embodiment, the scanner (1) has a ci of one pixel.
Light level level 0 (white) to level 7 (m) is threshold 1
If it is greater than or equal to 6, it is binarized and output as "lo" (black), and if it is smaller, it is binarized as "O" (white). In FIG. 2, one square forms one pixel, and the number inside the square represents the threshold value of that pixel.

Figure 3 (A), (B), and (C) are white, black, and intermediate colors.
FIG. 3 is an example of an image binarized using a threshold value according to the dither pattern shown in the figure. Figure 3 (A) is white (level 0), Figure 3 (B)
is black (level 7), Figure 3 (C) is intermediate color (level 4)
This is an example of a binarized image, and when the density level is between 1 and 5, even if the density of the input image is uniform, the 2-1 output will be white, as shown in Figure 3(C). This creates a pattern in which black appears alternately (a so-called checkered pattern).

Figure 4 is an example of a figure read by this device. As shown in figure
), the gray level changes from TRI ow to T.
Use the one between Hhiqh.

When you input a drawing with the text colored in this way into the scanner (+), the 3134th (A), (B), (C)
As shown in , the black line is black, the white background is white, and the colored part is in a checkered pattern.The binarized image is buffered (
2]. FIG. 6 shows an example of an image binarized into a checkered pattern of colored parts.

The reduction unit (3) reduces the image contents in the buffer (2) vertically and horizontally, respectively. This vertical and horizontal reduction can be achieved by dividing the entire image into 2×2 blocks of 4 pixels as shown in FIG. 7, and replacing each pixel block with one pixel.

This reduction unit (3) performs the above replacement using the following reduction method (A),
The reduction result (shown in Figure 8) of reduction method (A) is stored in the buffer (4), and the reduction method (B) is used.
The reduction result (shown in FIG. 9) is stored in the buffer (5). The above reduction method (A) calculates the logical sum of pixels A and tj among the four pixels shown in FIG.

Reduction method CB) is 31 pixels out of 4 pixels in Figure 9; g
Take the logical sum of C.

With this method, the image is reduced in two ways, but as shown in Figure 10, the intermediate color checkered pattern part becomes completely black in the reduction method (A), and the entire image in the reduction method (B), as shown in Figure 10. Becomes white. Therefore, in the case of the image illustrated in Fig. 6, the shape color part becomes completely black in the reduction method (A), and
), only the character component part in the shape color area is reduced as black pixels. FIG. 11 shows an example of images stored in buffers (4) and (5) in this manner.

The detection unit (6) processes the contents of the buffer (4) and detects the colored portion applied to the characters/symbols. This is, for example, a force that scans the contents of a buffer (0) vertically and horizontally, and if the number of consecutive black pixels (run length) is smaller than a threshold N, the consecutive black pixels are erased.
If the threshold value N is set to about 2 to 3 times the expected line thickness, only the colored portion remains as the processing result (see FIG. 12).

The remaining colored part image in this way is moved to X, buffer (5)
Assuming that the content of is Y, as shown in Figure 13, the logical sum XUY of image X and image Y becomes a colored character/symbol, and the logical sum XUY of one image . The character/symbol recognition unit (7) processes the xUY image, and the 0 graphic processing unit (8), which recognizes characters and symbols, processes the xUY image and recognizes the graphic part.

The general processing unit (9) includes a symbol recognition unit (7), a graphic processing unit (
Final recognition is performed by integrating the outputs of 8).

In the above embodiment, the dither pattern of scanner (1) 2 ((i conversion agony)/i was set so that the intermediate color binarization result would be a checkerboard pattern. ), the binarization result may be set to have a striped pattern.

In addition, in the above embodiment, the dither pattern of the binarization threshold is set to H/- in the scanner, and the result of the binarization by the scanner is stored in the buffer. However, if there is sufficient buffer capacity, the scanner can (It is also possible to input a multivalued image without converting it to fi, and then convert it into a binary image using S/W in later processing.

In the above embodiment, separation of dimensional areas separate from figures was dealt with, but as shown in Fig. 15, circuit symbols of electric circuits are colored to separate memorabilia elements in contact with connection lines. It can also be applied to.

[Effects of the Invention] As described above, according to the present invention, a drawing with colored text/symbols is input using a scanner with a binarization threshold set by a dither pattern, and the input image is converted to the one with the larger threshold of the dither pattern. By selecting the smaller one and processing it with two reduced patterns, we have adopted a method for separating and recognizing the character φ symbol and figure part in the drawing, making it easy to separate characters/symbols and figures. It has a perceivable effect.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of the present invention: J
The SZ diagram is the dither pattern of the scanner (1) in Figure 1, and Figures 3 (A), (B), and (C) are the binarization of white, black, and intermediate colors obtained with the dither pattern in Figure 2 above. Image pattern, Figure 4 is a table of figures to be input, Figure 5 is a table of figures after coloring, Figure 6 is an image pattern of the colored part, Figure 7 is a reduced image pattern, Figures 8 and Figure 9 is a table diagram based on reduction methods (A) and (B), Figure 10 is a reduced table diagram of an intermediate color pattern, Figures 11 (A) and (B) are each diagram after reduction, and Figures 12 and 13. Figures (A) and (B) are respective tables of figures after minutes, Figures 1 (a) and (b), Figure 15 is an explanatory diagram of another example, and Figures 18 (A) and CB. ) is an explanatory diagram of the conventional method. In the figure, (1) is a scanner, (2), (4), and (5) are buffers, (3) is a reduction unit, and (6) is a detection unit. (7) is the character e symbol recognition unit. (8) is a graphic processing section, (9) is a general processing section, and (10) is a main control section. In each figure, the same reference numerals indicate the same or corresponding parts. Agent Masuo Oiwa Figure 1 Figure 2 Figure 3 (?
(2-1 [B-B) Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Figure 12 Figure 13 (B) (A) Procedural amendment (voluntary) Showa year, month, day 2, name of the invention character/symbol separation reading method 3, representative of the person making the amendment Moriya Shiki 4, representative specification, claim 17 column, and the: Y details of the invention Explanation 6, Contents of Supplementary IF (1) The claims of Specification-1 are amended as shown in the attached sheet. (2) On page 9, line 17 of the specification, only the colored portion remains as a result of the processing (see Figure 12). '' is supplemented with ``If the processing result is an image in which only the colored parts remain and the broken line figures disappear due to the processing (see Figure 2).'' 7. Attachment 3 [1st Record Supplement iE? Page 1 of 11 stating the claims of tt.
Supplementary Part II - Patent No. 11 [Claims are stated]. -1 page (
1) Among written elements such as page 1 and chart diagrams, from L to ゛r of the description of specific letters and symbols, this ゛r has letters and symbols that are transparently colored. drawing,
A scanner is used to read a chart, etc. using a dither pattern in which a plurality of large and small numbers of binarized values are arranged in a matrix.
Create a reduced image by selecting/V the larger value of the digitized value, and select the smaller value of the binary value indicated by - and reduce it to create a reduced image, and then perform the above steps. A colored area corresponding to the above characters/symbols is extracted from the reduced image with the smaller binarization threshold value, and the above extracted area is converted into the characters/symbols from the reduced image with the larger binarization threshold value. Processed as
A separate reading method for characters and symbols characterized by a structure in which the outside of the face extraction area is processed as a figure. (2) Does the above scanner use pixels with large and small values of the binarization threshold? 2. The character/symbol separation reading system according to claim 1, wherein the system is configured by setting a dither pattern in which 11- positions are arranged in an intersecting t1- pattern.

Claims (2)

[Claims] (1) Among written elements such as drawings and chart diagrams, translucent coloring is applied over the description of specific characters and symbols, and drawings, chart diagrams, etc. that have characters and symbols with this translucent color are The image is read by a scanner that is set in a dither pattern with multiple binarization threshold values arranged in a matrix, and from the image read by this scanner, the one with the larger binarization threshold value is selected and reduced. Along with creating a reduced image,
Select the smaller value of the binarization threshold value and reduce it to create a reduced image, extract the colored area applied to the above characters/symbols from the reduced image with the smaller binarization threshold value, and then binarize it. The above extraction area of the reduced image with the larger threshold value is
A separate reading system for characters and symbols, characterized in that the characters and symbols are processed as symbols, and the areas outside the extraction area are processed as figures. (2) The scanner is configured by setting a dither pattern in which pixel units of large values and small values of the binarization threshold are arranged alternately. Separate reading method for characters and symbols.