JPH0514952B2 - - Google Patents

Description

[Detailed Description of the Invention] (Field to which the invention pertains) The present invention directly reads characters on a document with a machine.
In OCR (Optical Dharacter Recognition) equipment, in addition to characters arranged at a nearly constant pitch,
The present invention relates to a character cutting device capable of efficiently cutting out individual characters, symbols, etc. even from a character string containing successive alphanumeric characters, symbols, etc. having a pitch comparable to that of ordinary characters.

(Prior art) In the conventional method, when alphanumeric characters, symbols, etc. with a pitch of about half the normal character pitch appear consecutively in a character string, several of these alphanumeric characters and symbols are combined into one. Because it is cut out as a character,
There was a drawback that the cutting process was not performed properly. In addition, a method is being considered in which the information from the character recognition device is used to re-extract characters that have become rejects, but since that method requires recognition processing for all characters, the capacity of the recognition dictionary is limited. However, there are disadvantages in that the character cutting device becomes large, or the entire system becomes large as a character cutting device.

(Object of the Invention) In order to solve these drawbacks, the present invention provides a dictionary for recognizing alphanumeric characters, symbols, etc. that have a pitch different from normal characters, or only characters that have a different pitch and separate characters. , extract only those shapes that can be separated into multiple sub-figures from among the figures once cut out as characters, apply recognition processing to each sub-figure, and determine whether or not to be rejected; Based on the results, it is determined whether a figure cut out as a character can be separated into a plurality of alphanumeric characters, symbols, etc.

(Structure and operation of the invention) FIG. 1 is a block diagram showing the structure of an embodiment of the invention, in which 1 reads and stores image data of a character string extracted from a document image using a signal line a. character string image data storage device, 2 is signal line b
a character primary cutting device which reads image data of a character string stored in a character string image data storage device 1 using a signal line c based on an estimated pitch inputted using a signal line c, and performs primary cutting of a character; 3 uses the signal line c to convert the image data of the character string stored in the character string image data storage device 1 based on the primary character extraction result inputted from the primary character extraction device 2 using the signal line d. This is a secondary character extraction device that performs the necessary character recognition processing using the standard character pattern read in from the identification processing dictionary 4 through the signal line f, and corrects the primary character extraction result. The secondary cutting result is output using the signal line g. Each device will be explained below.

The character string image data storage device 1 is a normal OCR,
Alternatively, it is an image memory that stores image data for one character string extracted by a separately pending patent application No. 55-126845 "Two-dimensional Character Area Extraction Apparatus" or the like.
The primary character cutting device 2 is a device that cuts out individual characters based on the estimated character pitch input from the outside.
For example, this can be realized by the patent application No. 56-74015 "Character cutting device" which has been filed separately. Here, you can enter a numerical value directly for the estimated character pitch, or
For example, it can be easily estimated by projecting a character string in the column direction and finding the width of the character string, taking advantage of the property that the ratio of height to width of kanji is approximately 1.

FIG. 2 is an explanatory diagram of the character string image array and the results extracted by the primary and secondary character extraction devices.
1 shows a character string image array, and 2 and 3 show examples of extraction results of the primary character cutting device 2 and the secondary character cutting device 3.

In FIG. 2, the numbers "34" and ")." in FIG. 2 are extracted as one character. Originally, these characters should be cut out in two parts, but
Individual numbers and symbols are usually only about half the size of letters, and because they exist consecutively, two numbers and symbols have been cut out together.
This is difficult to handle with conventional OCR. On the other hand, the character "mon" is accurately extracted by using a predetermined estimated character pitch.

The secondary character cutting device 3 is the primary character cutting device 2.
Read the contents of the text, extract figures that can be separated into multiple partial figures, cut out the figures as characters, perform recognition processing on each partial figure cut out as characters, or the whole figure and each partial figure, and cut out as characters. This is a device that determines whether a drawn figure is a single character as it is, or whether it can be broken down into multiple figures. To determine whether characters are separated, for example, the vertical direction of the character string (in the case of horizontal writing),
Or perform horizontal projection (in case of vertical writing),
It is only necessary to check whether there is a part in the center of the figure extracted as a character by the primary character extraction device 2, in which no black pixels exist.

FIG. 2 shows the result of cutting out by the secondary character cutting device 2, where “〓” and “〓” are rejected as a result of identification, and “〓” and “〓” are rejected as a result of identification, and the “gate” of the cutting result of the primary character cutting device 2 is shown. ” has been adopted as is.
Further, "3", "4", ")", and "." are recognized as numbers and symbols, and the results of the single character cutting device 2 are corrected and cut out as separate characters.

The recognition processing dictionary 4 is used to register standard patterns of characters used during the recognition processing of the secondary character cutting device 3, and only standard patterns for characters whose pitch is different from normal characters such as alphanumeric characters and symbols are registered. Therefore, the dictionary capacity can be made extremely small. In addition, by registering standard patterns for separated characters such as "mon", it is possible to improve accuracy by performing recognition processing not only for individual partial figures but also for the entire separated character figure. . In this case, there are few characters in which the shapes that make up the characters are completely separated, such as "mon" and "cho", and even in this case, the dictionary capacity is much smaller than having a standard pattern for the entire character. . Further, the standard patterns registered in the recognition processing dictionary 4 may be written in advance in a ROM etc.
Also, standard patterns may be learned through conversation processing with humans during the extraction process.

(Effects) As explained above, this device extracts only the separated figures among the characters cut out from the character string, and recognizes the separated individual partial figures, individual partial figures, and the whole figure. By performing this processing, individual characters can be efficiently cut out from character strings that include not only regular characters that are lined up at a constant pitch, but also consecutive alphanumeric characters that have a pitch that is only about half that of regular characters. It also has the advantage that the dictionary required for recognition processing can be extremely small. Furthermore, it is clear that the method used in this device is fully applicable not only to the case of modifying the result of cutting once, but also to the process of cutting out characters from the end of a character string.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention, and FIG. 2 is an explanatory diagram of a character string image arrangement and the results extracted by the primary and secondary character cutting devices. 1...Character string image data storage device, 2...Character primary cutting device, 3...Character secondary cutting device, 4
... Dictionary for recognition processing, a ... Signal line for transferring character string image data, b ... Signal line for inputting the estimated character pitch necessary for primary character extraction, c ...
Signal line for the primary character extraction device to read character string image data, d... Signal line for outputting the primary character extraction result, e... Signal line for the secondary character extraction device to read the character string image data. signal line, f
...Signal line for the secondary character cutting device to read the dictionary for recognition processing, g...Output signal line.

Claims (1)

[Claims] 1. A character string image data storage device that inputs image data of a character string part in a document image and stores an image of the character string, and uses the character string image data stored in the character string image data storage device. A primary character extraction device performs character extraction based on estimated character pitches input from the outside. or extract only those that can be decomposed into more than 10 sub-figures, and perform recognition processing on each sub-figure or each sub-figure and the entire figure,
a secondary character cutting device that determines whether to cut out the figure cut out by the primary character cutting device as a single character or as a plurality of characters;
It is equipped with a recognition processing dictionary that stores standard patterns of characters used during recognition processing by the secondary character extraction device, and in addition to regular characters that are lined up at approximately regular pitches, it also has English characters that have a pitch that is about half that of regular characters. A character cutting device is characterized in that it cuts out individual characters from a string of consecutive numbers, symbols, etc.