JPS60195685A - Character segmenting circuit - Google Patents

Abstract

PURPOSE:To obtain a character segmenting circuit which can reduce a memory capacity, by calculating a density gradient direction histogram in the direction of character line and detecting the direction of a character string by collating the calculated histogram with the direction histogram of a standard pattern. CONSTITUTION:A picture memory 1 stores a picture to be processed immediately out of read pictures and differentiation arithmetic circuits 5 and 6 calculate and detect the horizontal and vertical components of the density gradient vector of stored pictures. A density gradient direction detecting circuit 7 detects the direction of the density gradient vector. A direction histogram calculating circuit 8 calculates the direction histogram of density gradient vectors from the direction of the vectors obtained for each picture element and causes a histogram memory 10 to record the histogram. A direction detecting circuit 11 compares the standard histogram of another histogram memory 9 with the direction histogram and detects the direction of a character string to be processed. A projection circuit 12 projects the picture against the axis normal to the detected direction of the character string and a line separating circuit 13 finds the minimum value, 2nd mean value and 3rd mean value of the projected values by scanning the projection.

Description

Detailed Description of the Invention (a) Technical Field of the Invention The present invention detects the direction of one character string by using a concentration gradient direction histogram showing the distribution status of two character line directions, and then separates two lines. The present invention relates to a character extraction circuit that performs character extraction using two-stage character separation processing.

(b) Technology background As an image processing method for recognizing figures and objects.

Optical Image Processing 5 Various methods such as electrical analog processing and digital image processing have been considered and are being studied. Among these, digital image processing using a computer or a dedicated processing device has the following characteristics: non-linear processing can be easily performed, processing and parameters can be changed by a program, making it highly flexible, and high accuracy. The field of application is wide-ranging, and there are great expectations for the future as industrial image processing such as medical image processing, non-destructive inspection, visual inspection, and article recognition.

While digital image processing has the above characteristics,
There are still various shortcomings and undeveloped areas.

There is a demand for the development of image processing technology that has a wider range of applications and is more reliable.

(c) Prior Art and Problems Next, a character extraction method, which is a main component of image processing technology, will be explained with reference to one drawing.

FIG. 1 shows an outline of a conventional character extraction method.

In the figure, 1 and 2 are image memories, and 3 is a positioning circuit 74, which is a character cutting circuit. Also, ■ indicates input image data.

FIG. 1 shows an image memory 1 for recording input image data. Image memory lJ2 that stores standard image patterns. A positioning circuit 31 that aligns the image data recorded in the image memory 1 with the standard image pattern stored in the image memory 2; a character cutting circuit that cuts out the image data aligned and corrected by the positioning circuit 3 into individual characters; It consists of 4.

The conventional character cutting method shown in FIG.
Correct the entire position of the string.

This is a method of cutting out each character by referring to the position of each character in a standard image pattern stored in advance.

However, this method has disadvantages in that it takes a long time to process the alignment, and that a large capacity image memory 2 is required since it is necessary to store a larger number of standard image patterns.

(d) Purpose of the Invention The object of the present invention is to provide a novel character extraction circuit that eliminates the above drawbacks, and particularly to realize a character extraction circuit that can reduce memory capacity and speed up character extraction processing. be.

(e) Structure of the Invention An object of the present invention is to provide a character cutting circuit that cuts out individual characters from a character string consisting of one or more lines, and which shows the distribution situation in the character line direction on the character string. a first means for calculating a density gradient direction histogram; a second means for detecting the direction of the character string by comparing the direction histogram calculated by the first means with the direction histogram of a standard pattern; a third means for projecting an image of the character string onto an axis perpendicular to the direction detected by the second means; fourth means for detecting a position and separating the character string into individual lines; and aligning the individual lines separated by the fourth means on an axis parallel to the direction detected by the second means. and a sixth means for detecting the position of each character from the projected image of each line obtained by the fifth means and separating each line into individual characters. .

This can be achieved by providing a character extraction circuit that is characterized by reducing memory capacity and increasing the speed of character extraction processing.

(f) Examples of the invention The present invention will be explained below with reference to the drawings.

FIG. 2 shows an embodiment of the character cutting circuit according to the present invention, FIG. 3 shows a projected view of the character cutting circuit according to the present invention, and FIG. 4 shows a character cutting diagram according to the present invention.

In the figure, 5 and 6 are differential operation circuits, 7 is a concentration gradient direction detection circuit, 8 is a direction histogram calculation circuit, 9.10 is a histogram memory, 11 is a direction detection circuit, and 12.14
13 represents a projection circuit, 13 represents a line separation circuit, and 15 represents a character separation circuit.

Note that the same symbols as in Figure 1 indicate the same content, and also in Figures 3 and 4.
In the figure, R1 to R3 indicate rows of character strings, and Tl to Tl0 indicate individual characters, respectively.

This embodiment uses the image memory j described in FIG.

a differential calculation circuit 5 for detecting the horizontal component 3 and the vertical component of the density gradient vector of the image data stored in the image memory 1;
61. A density gradient direction detection circuit that detects the direction of the density gradient vector of image data. 71. A direction histogram calculation circuit that calculates a histogram from the direction of the density gradient vector obtained for each pixel of the image data. Histogram memory 9 for storing the standard pattern histogram 1. Histogram memory 10 that stores the direction histogram of the image data to be processed calculated by the direction histogram calculation circuit 8, the storage contents of the histogram memory 9 and the histogram memory 1
A direction detection circuit 11 that detects the direction of the character string of the image data to be processed by comparing the stored contents of 0. A projection diagram in which image data is projected onto an axis perpendicular to the direction of the character string detected by the direction detection circuit 11! l 2. A row separation circuit 13 that scans the one-dimensional projected image obtained by the projection circuit 12, finds projection values in a predetermined section, and detects and separates sections in which the projection value is larger than a predetermined value.
．． A projection circuit 14 that extracts a plurality of regions from the outputs of the direction detection circuit 11 and the line separation circuit 13 and projects each region onto an axis parallel to the direction of the character string. The character separation circuit 15 scans each projection image obtained from the projection circuit 14, detects and separates regions having values larger than a predetermined projection value, and represents the positions of individual characters.

Next, the operation of this embodiment will be explained. It is assumed that the image memory l in this embodiment has a capacity of 512 x 512 xabits.

The image memory 1 stores an image to be processed for the time being out of the read input image data (2). Differential calculation circuit 5 calculates the horizontal and vertical components of the density gradient vector (vector obtained by analyzing the image in terms of density distribution) of the stored image.
, 6 to detect the calculation.

A density gradient direction detection circuit 7 detects the direction of the density gradient vector based on the detection data.

Further, the direction histogram calculation circuit 8 calculates a histogram from the direction of the density gradient vector obtained for each pixel, and records it in the histogram memory 10. The direction detection circuit 11 compares the standard histogram stored in the hisdogsim memory 9 storing standard pattern histograms with the direction histogram recorded in the histogram memory 10, and determines the character string to be processed (see FIG. , the direction of the uppercase alphabetic character string 1?l, R2, R3) shown in FIG.

Next, an image is projected by the projection circuit 12 onto an axis perpendicular to the direction of the detected character string (the projection situation is shown in FIG. 3(A)). The row separation circuit 13 uses the 1 obtained by the projection circuit 12.
The next projected image is scanned, and first the average value n of the minimum value, second smallest value, and third smallest value of the projection values in a predetermined constant interval from the scanning start point is calculated.

Next, the sum of the predetermined value t and the average value n is T11
As shown in FIG. 4(A), the sections (PI, Ll), (F2, 1.2) whose projection values are larger than T11 are
) ... (Fk, Lk) are detected separately. Output separated and detected by the row separation circuit 13 [section (PI, Ll)
, (F2.L2)... (Fk,Lk)) and the direction of the character string detected by the direction detection circuit 11, the image is divided into R1, R2,...Rk regions, and each The area is projected by the projection circuit 14 parallel to the direction of the character string as shown in FIG.
The projected images (K projected images) shown in B) are obtained.

The projected images obtained by the projection circuit 14 are transferred to the character separation circuit 1.
5, and find the average value mi (+=1 to k) of the minimum value, second smallest value, and third smallest value of the projection value in a predetermined certain section from the scanning start point.
Next, the sum of the predetermined value t and the average value mi is TH
i' (i=1 to k), and the interval (F?, Pi) in which the projection value shown in FIG. 4(B) is larger than Tl1i'
, <F:, si', > ... (pi, ti) are separated and detected. These sections represent the position of the characters in [(.

(g) Effects of the Invention According to the present invention as described above, only the directional histogram of the standard pattern, which requires only a small amount of storage, needs to be stored in advance, so the memory capacity can be reduced and high-speed processing can be achieved. This has the effect of providing a possible character extraction circuit.

[Brief explanation of drawings]

FIG. 1 is an overview of a conventional character extraction method, FIG. 2 is an embodiment of a character extraction circuit according to the present invention, FIG. 3 is a projection diagram of character extraction according to the present invention, and FIG. 4 is a diagram according to the present invention. Such character cutout diagrams are shown respectively. In the figure, 1.2 is an image memory, 3 is an alignment circuit, 4 is a character cutting circuit, 5.6 is a differential calculation circuit, 7 is a density gradient direction detection circuit, 8 is a direction histogram calculation circuit,
9. 10 is a histogram memory, 11 is a direction detection circuit,
12. 14 indicates the projection circuit, 13 indicates the line separation circuit, and 15 indicates the character separation circuit.

Claims (1)

[Claims] A character cutting circuit for cutting out individual characters from a character string consisting of one or more lines, the first means for calculating a density gradient direction histogram indicating the distribution situation in the character line direction on the character string; , a second means for detecting the direction of the character string by comparing the direction histogram calculated by the first means with a direction histogram of a standard pattern; and a second means perpendicular to the direction detected by the second means. a third means for projecting an image of the character string onto an axis; and detecting the position of each line of the character string from the projected image obtained by the third means, and and a fourth means for separating the individual rows separated by the fourth means into the second rows.
a fifth means for projecting onto an axis parallel to the direction detected by the means; and a fifth means for detecting the individual character positions from the projected images of the individual lines obtained by the fifth means, and A character cutting circuit characterized in that a sixth means for separating characters into characters is provided.