JPS58165178A - Character reader - Google Patents

Abstract

PURPOSE:To extract the quantity of an average ruggedness of a character pattern as a feature, by obtaining the length of the border line between a character line part and a background part for every circumscribed side in the peripheral part of the character pattern and obtaining the quantity of the average ruggedness of the character pattern on a basis of the relative value of this length of the border line to the length of a circumscribed frame. CONSTITUTION:The circumscribed frame surrounding the character pattern is obtained by detecting a left circumscribed side L which is brought into contact with leftmost end, a right circumscribed side R which is brought into contact with the rightmost end, an upper circumscribed side U which is brought into contact with the uppermost end, and a lower circumscribed side D which is brought into contact with the lowermost end. Circumscribed sides U, D, L, and R are used as reference positions to scan the character pattern from respective circumscribed sides U, D, L, and R toward the center of the character pattern, and length of scanning line to character line is obtained. The total sum of differences between these scanning lines is calculated for each of circumscribed sides U, D, L, and R, and the calculated value is detected as the length of the border line between the background part and the character line part of the character pattern. Thus, feature required for every rough classification of original character pattern of hand-written character is extracted stably and efficiently.

Description

[Detailed Description of the Invention] [Technical Portion of the Invention] The present invention provides a highly practical character reader that can stably and efficiently extract features for broad classification identification from a large number of characters including kanji. Regarding equipment.

[Technical background of the invention]

Extract the image of the input character pattern, match it with the features of the dictionary pattern registered in advance in the dictionary, and read the input character pattern according to the result of the match. .

However, when a large number of characters, including Chinese characters, are to be read in the character iin, the amount of processing required to match the above-mentioned characteristics becomes extremely large, and the processing efficiency deteriorates considerably. Conventionally, prior to the above recognition process, the characters to be recognized are roughly classified according to multiple general special instructions, which reduces the number of characters to be matched at the identification stage, improving processing efficiency and speeding up processing speed. There are many things being done. At this time, it is necessary to use the same features for the above-mentioned major classification and the 4I features used for candidate character identification, but this 4111 is effective against deformation of characters and patterns caused by handwritten characters, etc., and various noises. It is necessary to be sufficiently stable against the

However, conventionally, the features used to broadly classify character patterns have focused on the complexity of the character lines, the shape of the peripheral part of the character, and even the rough shape of the entire character. For example, the following literature introduces these in detail.

Sakai, Watanabe 1 Current status of printed kanji recognition l.

Information Processing, Vol, 22 Ma4 P.72'14-2
79 (April 1982) Figure 1 shows an example of this. Focusing on the shape around the original character pattern, scanning areas U, D, L, and R are provided in rectangular shapes on the top, bottom, left, and right of the periphery. Detects the character line that exists within and sets its amount to "0".

It is quantized into three levels rlJ and r2J and used as feature data. What is shown in Fig. 2 focuses on the circumscribed sides u, d, j, and r that touch the original character pattern, calculates the area of the character background that touches these circumscribed sides u, d, I, and r, and calculates this by is used as a multidimensional feature vector. The difference is that the former (Figure 1) focuses on the character line area, while the latter (Figure 2) focuses on the area of the background area, but both are characterized by the area quantity around the character pattern. It can be said that this is true.

[Kanden point of background technology]

11. The mold extraction mentioned above in 4I is very effective when the pattern is standardized, such as in printed characters. : is unstable in the case of characters such as handwritten characters whose characteristics as characters are significantly deformed from the original characters. Moreover, even if the characters are printed, if the character pattern is chipped or faded, feature extraction becomes unstable. In other words, there is a running time in which it is not always possible to perform stable image extraction against character patterns and various types of noise.

[Purpose of the invention]

The present invention has been made in consideration of these circumstances, and its purpose is to read a large number of characters, including handwritten kanji, by reading the original character pattern written by hand. It is an object of the present invention to provide a highly practical character reading device that can stably and efficiently extract necessary features.

[Summary of the invention]

The present invention provides classification information (41 characters) necessary for character classification identification.
The average of the character pattern is calculated from the relative value of the length of these border lines with respect to the length of the circumscribed frame. Find the amount of unevenness and use this as major classification feature 2
The present invention relates to a character reading device that extracts characters.

〔Effect of the invention〕

According to the present invention, a novel concept of a boundary ls- between a character line portion and a background portion corresponding to each circumscribed side of the character pattern, which is slowed down by the circumscribed side touching the input character pattern,
In other words, the degree of unevenness around the character, in other words, the complexity of the shape, is introduced, and the relative size of these boundary line lengths with respect to the circumscribed frame is characterized, so the change in the size of the character pattern Characteristic information can be obtained sufficiently stably against various noises such as errors and positional deviations, and the broad classification of input patterns can be stably and efficiently identified.
A significant improvement in cognitive processing can be achieved. In addition, even if you are dealing with complex character patterns such as kanji, handwritten characters, or even printed characters, you can use the simple method described below. Since it can be easily obtained, it can have a remarkable effect of high realism in character recognition processing.

[Embodiments of the invention]

Hereinafter, one embodiment of the present invention will be described with reference to the drawings.

FIG. 3 shows the feature extraction process of the input character pattern in this apparatus, and here the character pattern "天" input by hand is shown. The feature extraction process is
First, for a given character nosetan, a circumscribing frame surrounding it is detected. For this circumscribing frame, inspect the circumscribed edge L1 that touches the leftmost edge of the character pattern, the right circumscribed edge R1 that also touches the rightmost edge, the rising tangent U that touches the top edge of the character pattern, and the lower circumscription mD that touches the bottom edge. It is determined by helmet, each of the above circumscribed sides U,
D, L, and R can be easily determined by scanning the character pattern and calculating the limit values of the position coordinates of the character lines.

Each circumscribed side U, with R as the reference position (reference line).

Scanning is performed from D, L, and R in the direction toward the core of the character pattern, and the length of the scanning line until reaching the character line is determined. However, from the viewpoint of signal processing, it is preferable that the maximum flame of the above-mentioned scanning be set at a constant ratio depending on the size of the character pattern. Then, the sum of the differences between these scanning line lengths is calculated as each circumscribed side U, D, L, H.
, and this value is detected as the boundary line length between the background part and the character line part of the character pattern. For example, the length LL of the boundary line around the circumference is determined by defining the maximum flame as hH (h<1) as shown in Figure 3, and scanning in the direction toward the center of the character pattern with L as the reference position. Find the length.

However, when the scanning line did not reach the character line, the length hH was reached. I promise to finish the scan. Then, for the scanning lines obtained in this manner, the difference in length (absolute, relative value) between adjacent scanning lines is added as shown in FIG. 4, and then the vertical width V of the character is added. 1 The contour line length lL (the length of the thick line in Fig. 4) that touches the circumference obtained in this way is stabilized against the width of the character line of the character pattern, the inclination of the character, and the positional shift, and covers a wide range. It is a feature quantity. Then, by similar processing, the contour line lengths of the background portions that are in contact with the other circumscribed sides R, U, and D, respectively, are determined.

Then, use these values as the size of the circumscribing frame of the character pattern,
In other words, the boundary line length normalized with respect to the vertical and horizontal widths is determined as □. However, in the above formula H
, V indicate the horizontal and vertical widths of the character pattern circumscribing frame. As a result, information indicating the characteristics of the surrounding area of the input character pattern or a set of boundary line lengths (CL* C”eCυ
, C shift). This information corresponds to the average degree of unevenness of the outer shape seen from the circumscribed sides U, D, L, and R of the character pattern, or in other words, the complexity of the surrounding area of the character, and therefore, the character pattern can be broadly classified. This sufficiently reflects the characteristics of the character pattern.

That is, by using the characteristics (CL, Cm, Ctr, CD) of the surrounding area of the character pattern obtained in this way, it becomes possible to effectively, stably and reliably classify and identify the character pattern. . - For example, when the handwritten characters ``古'' in various shapes as shown in Figure S (a), (b), and (C) are input, the boundary of the upper side of the character pattern is Line length et, t, 2. , ts, these are all 1 of the pattern of the shape shown in Fig. 5(d).
．． This corresponds to counting (=Ha +hv,). Therefore, if these feature information are normalized by the character size, the feature values will be the same as long as the ratio of H and ■ is the same.

That is, it can be said that (CL, CR, CU, CD) are features that are stabilized against character deformation.

FIG. 6 is a schematic diagram showing an apparatus according to an embodiment of the present invention, which performs such processing to extract features for broad classification identification of character patterns.

A character pattern 1 to be read is photoelectrically converted by a scanning-photoelectric conversion device 2, which is made up of, for example, a television camera, and is input by scanning the character surface. The image signal of the character pattern 1 inputted through the photoelectric conversion device 2 is inputted to the binary quantization device 3, and is discriminated at a discrimination level determined based on the background level, for example, and converted into binary quantization. be done.

The quantized character pattern signal is then stored in a pattern storage device fi114 such as a one-frame memory as a binary pixel signal corresponding to the scanning position. The literary pattern pixel signals stored in the storage device 4 are scanned in a predetermined direction and read out, and are subjected to the feature extraction process described above.

That is, the circumscribing frame detection device 1 searches for the positional coordinates of the character line of the character pattern stored in the storage device 4, finds its leftmost end m*xL, its oldest end position llll , mXL, let the extra circumscribed tangent R be Ra”X
I'm looking for it as a li.

At the same time, the uppermost coordinate YtF and the lowermost coordinate @l'D of the character line of the character pattern are determined, and the ascending tangent U and the lower circumscription are determined. Then, the circumscribing frame detection device 5 sends information about these circumscribing areas, R, U, and D to scanning circuits #L, 11 . ＃υ、
Each is given as control information to 6D.
In each of the scanning circuits gla, 6 islands, 6U, and 6D, a reference line for starting scanning is determined from the information on the given circumscribed side, and the signal of the character pattern stored in the storage device 4 is transferred from the reference line to the center of the character pattern. In the direction of (left → right), (right →
(left), (top → bottom), and (bottom → top) to find the scanning line length corresponding to (contacting) each circumscribed side. The scanning line obtained in this way! The information is from the boundary line calculation department FL.

1 island, 7 υ, and 1 school, and as mentioned above, calculate the sums 1b, In, lυ, and In of the scanning line length differences, respectively.
2) Normalize this to get the relative quantity C j'i:,
・、.・、C・is required"□゛・:.

This information is integrated and given to the feature comparison device I as feature information (CL, CB, Co, Co) around the input character pattern.

In the feature dictionary, the above-mentioned feature information for each character to be recognized is registered in advance, and the feature comparison device 8 compares these dictionary features with the features of the input character pattern obtained. The similarity (degree of similarity) between the two is calculated sequentially. Then, when the degree of similarity is greater than or equal to a predetermined tolerance value, the category of the feature dictionary obtained is outputted as a major classification identification result.

As described above, the embodiment of the apparatus for extracting the 4I image and performing the broad classification identification based on the extraction of the 4I image according to the present invention can be realized very easily. Then, by applying the thus obtained large classification identification result and the image signal of the input character pattern to the next stage character iim section, it becomes possible to easily and efficiently perform the character gm. Moreover, as mentioned above, it is stable against deformation of character patterns such as handwritten characters and noise, so recognition #! & It can significantly improve the efficiency of the process.

Therefore, it is possible to construct a highly practical character reading recognition system that recognizes many characters including Chinese characters, and has great effects.

However, the present invention is not limited to the embodiments described above. For example, in the embodiment, the boundary line length was determined as the sum of the length differences between adjacent scanning lines, but a similar effect may be obtained by directly tracing the outline of the character area to determine the boundary line length. . In this way, although the amount of data processing increases, it is possible to obtain more detailed and accurate border line lengths for character lines and background parts. Major classification identification may also be performed using feature information and the like. In short, the present invention can be implemented with various modifications without departing from its scope.

[Brief explanation of the drawing]

1 and 2 are diagrams for explaining the concept of conventional character pattern % image extraction, FIG. 3 is a diagram for explaining the concept of character pattern feature extraction according to the present invention, and Figure 4 is a diagram for explaining the concept of character pattern feature extraction according to the present invention. The figure is a schematic configuration diagram of the main parts of the character butterfly device according to the embodiment of the present invention. 1,... Character, pattern, 2... Scanning photoelectric conversion device, 3... Binary at conversion device, 4... Pattern memory attachment, 6... Circumscribing frame detection device, 6L, 6%, 6υ, 6D・
・・Scanning circuit, r L e F l e rυ, 71D
...Boundary line calculation unit, 8...Feature comparison device, Ri...
Feature dictionary. Applicant's agent: Patent attorney Takehiko Suzue 1 ・□). , :. Figure 1 Figure 2

Claims (2)

[Claims] (1) A manual that inputs the # character pattern by photoelectric conversion,
1. means for binary quantizing and arranging the input original character pattern; means for detecting a circumscribing frame consisting of circumscribing edges in four directions, top, bottom, left and right of the stored original character pattern; 1. Means for detecting the contour line of the character background part in the direction from each circumscribed side toward the center of the original character; A character reading device having a 4I image, comprising means for detecting an average amount of unevenness around an original character pattern, and means for broadly classifying the original character pattern from the detected average amount of unevenness. (2) The means for detecting the length of the contour line of the character background section scans from each circumscribed side toward the center of the original character pattern, and accumulates the difference in scanning distance from the circumscribed side to the character line. A character reading device according to claim 1.