JPS63249282A - Multifont printed character reader - Google Patents

Abstract

PURPOSE:To efficiently perform the document processing by severely collating one-page components of input characters at every font and defining the font of the character, which has the lowest reject (unreadable) rate, as the font of this page and collating one-page components of input characters with a relaxed reference with respect to only this font to output answers. CONSTITUTION:A collating means 2 calculates the degree of noncoincidence between feature data of feature data of an inputted character and that of contents of a dictionary, and an answer is outputted when the degree of noncoincidence is lower than a prescribed threshold and the difference between first and second candidates is larger than a prescribed threshold, but reject is outputted otherwise. A font determining means 3 counts the number of unreadable characters at every page of an input document with respect to each font of dictionaries 11, 12... and determines the font, where the number of unreadable characters is minimum, as the font of this page. A font selecting means 4 selects respective fonts at every page of the input document in the first stage and selects the font, which is determined by the font determining means 3, in the second stage. A threshold selecting means 5 selects a decision threshold level different between the first stage and the second stage. Thus, characters are efficiently read with high precision.

Description

[Detailed Description of the Invention] [Summary] This is a device that reads printed characters in various fonts, has dictionaries for each font, performs strict verification for each font against pages of input characters, and rejects (reads) characters. impossible)
The font with the lowest rate is regarded as the font for that page, and the answer is output by comparing that font alone using loose criteria.

[Industrial Field of Application] The present invention relates to a printed character reading device, and more particularly to a reading device for a document printed with characters of multiple types of fonts.

Literature, patent information, information magazines from foreign countries, etc. can be input into a computer as character code information and used for classification, organization, searching, high-speed reference, etc. Automatic translation, automatic distribution, and automatic design using computers are also possible. However, most of this information is obtained manually using a keyboard, which requires an enormous amount of manpower and time.

One way to solve this problem is to simply input these documents into a scanner, but there is currently insufficient technology to convert the documents into character codes and into understandable code information for a computer.

One of these problems is that the shapes of characters vary widely in each country and in each field, and there is a lack of technology that can immediately respond to these variations.There is a need for technology that can respond to this problem.

[Prior Art] In conventional printed character reading devices, an operator usually stores the font used in a document to be read and then reads the document.

There is an example of multi-font support for line printer printed characters that are only numbers, but the method is to prepare dictionaries for all fonts, store all dictionaries of the font to be read without dividing them by font, The answer indicated by the dictionary with the minimum degree of mismatch with the input pattern is output as is.

[Problems to be solved by the invention] The above conventional multi-font support technology was able to handle cases where the category was only numbers, but when the number of character types to be read expanded to various categories such as alphanumeric, kana, and symbols. The disadvantage is that it cannot be handled.

Of course, when using multiple fonts, even though they are in the same category, they have different letterforms, so they cannot be read with the same dictionary, which is equivalent to an increase in the number of categories, and there is a problem that highly accurate reading becomes impossible. .

The present invention aims to provide a novel multi-font printed character reading device that solves the problems of the prior art.

[Means for Solving the Problems] FIG. 1 shows a block diagram of the principle of a multi-font printed character reading device according to the present invention.

In the figure, 1+, 12.13 are dictionaries, which are stored for each font.

2 is a matching means that calculates the degree of mismatch between the input character feature data and the feature data of the dictionary, and determines if the degree of mismatch is less than a predetermined threshold and the difference between the first and third candidates is within a predetermined value. If it is larger than the threshold, output the answer, otherwise output unreadable.

3 is a font determining means, dictionary I 1.12.13
, -, the number of unreadable characters per page of the input document for each font is counted, and the font with the least number of unreadable characters is determined as the font for that page.

Reference numeral 4 denotes a font selection means, which selects each font for each page of the input document in the first stage, and selects the font determined by the font determining means 3 in the second stage.

Reference numeral 5 denotes a threshold selection means, which selects different judgment/threshold values in the first stage and the second stage.

[Operation] The multi-font printed character reading device of the present invention reads documents in which various fonts do not coexist within a page, but do coexist between pages.

Recognition Dictionary 1+, 12.13. −・～ are prepared for each font, and all dictionaries of the font to be read are prepared in each tone.
The font dictionary, patient 2 font dictionary, and N13 font dictionary are stored separately.

In the first stage, characters are extracted one by one from the image information input by the scanner, data to be compared with the dictionary is extracted, and then the data is compared with the dictionary for each font, and if the degree of mismatch is less than the first threshold, - The answer is output when the difference between the first place candidate and the third place candidate is greater than the second threshold, and is rendered unreadable otherwise.

After the answers for all the characters in one page are output as answers for each font, the font determining means 3 counts the number of unreadable characters in one page, compares the number of unreadable characters by each font, and selects the least number of unreadable characters. Decide the font as the reading font for that page.

In the second step, the dictionary of the font determined by the font determining means 3 is selected, the first true reading is performed again, the degree of mismatch is determined by comparing with the dictionary, and the first threshold value is set to a value smaller than this. , the second threshold is switched to a fourth threshold larger than this, and the degree of inconsistency is smaller than the third threshold, and the candidates are the first and second-value candidates. output the answer when the difference is greater than the fourth threshold,
Outputs the final answer as unreadable.

This makes it possible to read multi-font printed characters with high precision and a low rejection rate.

Of course, for documents for which the font to be used is clearly known, it is possible to set the verification font and start reading from the second stage.

[Example] The present invention will be described in more detail below with reference to Examples shown in FIGS. 2 to 5.

FIG. 2 is a block diagram showing the configuration of an embodiment of the present invention.

In the figure, 01 is a scanner that optically scans an input document and performs photoelectric conversion to obtain binary image data.

02 is an image memory, which stores output image data of the scanner O1.

11, 12.13 are dictionaries of font 1lhi, font 龘2, and ``font N113, respectively.

A control processor 21 cuts out characters from the image data in the image memory 02, extracts data to be checked against a dictionary, and controls the overall data flow.

22 is a collation circuit, which collates the collation data extracted by the control processor 21 and the data in the dictionary 11, 12, 13;
An answer or unreadable is output based on a threshold value selected from the threshold value memory 51 according to the recognition stage.

A work memory 31 temporarily stores collation data, first-stage answers, fonts, and the like.

61 is a memory that stores the output of the final answer.

3 to 5 are diagrams showing the flow of processing according to an embodiment of the present invention.

FIG. 3 is a diagram showing the procedure of the verification process in the first stage.

Characters are separated from one page of image data from the scanner, collation data is extracted, and the data is collated with the dictionary of each font, and the answers for one page of each font are stored in the work memory.

The operation will be described below with reference to the numbers assigned to the data flows shown in FIG.

One page of image data ■ from the scanner is stored in the image memory.

The control processor retrieves the image data from the image memory ■, separates the characters, extracts the matching data, and temporarily stores it in the work memory ■.

The control processor reads the first threshold value and the second threshold value from the threshold memory and sets them in the verification circuit.

The collation circuit collates one page of collation data in the work memory with the dictionary data ■ of font N[11, and temporarily stores the answer in the work memory.

Next, it is compared with the dictionary data ■ of Font Arashi 2, and the answer is temporarily stored in the work memory. Next, it is compared with the dictionary data ■ of font N[L3, and the answer is temporarily stored in the work memory.

FIG. 4 is a diagram showing the procedure of font determination processing.

The control processor counts the number of rejected characters from the answers for one page of each font, compares the number of rejected characters of each font, and determines the one with the smallest number of rejected characters as the reading font.

FIG. 5 is a diagram illustrating the comparison between the first stage (when determining the font) and the second stage (when outputting the final answer) in this embodiment.

The threshold value is set so that the matching rejection rate is high and the misreading rate is low when determining the font, and the matching when outputting the final answer is set so as to keep the rejection rate as low as possible.

[Effects of the Invention] As explained above, according to the present invention, printed characters in which various fonts are not mixed within a page but mixed between pages can be read with high accuracy and with a low rejection rate, which improves document processing. The effect on efficiency is extremely large.

[Brief explanation of the drawing]

Fig. 1 is a block diagram of the principle of the present invention; Fig. 2 is a block diagram showing the configuration of an embodiment of the present invention; Fig. 3 is a diagram showing the procedure of the first stage matching process in an embodiment of the present invention. , FIG. 4 is a diagram showing the procedure of font determination in one embodiment of the present invention, and FIG. 5 is a diagram explaining the collation between the first stage and the second stage. In the drawings, 1 r, 12.11, 11.12.13 are dictionaries, 2 is collation means, 3 is font determination means, 4 is font selection means, 5 is threshold selection means, 01 is scanner, 02 is image Memory, 21 is a control processor, 22 is a collation circuit, 31 is a work memory,
51 is a threshold memory, and 61 is an answer output memory. Principle block diagram of the present invention Fig. 1 Fig. 2 Fig. 3 Font determination Selection (cutting) Explanation of matching between the first and second stages Whistle 5 Figure

Claims (1)

[Claims] Dictionaries stored for each font (l_1, l_2, l_3
), a matching means (2) that calculates the degree of mismatch between the input character feature data and the dictionary feature data, and outputs the answer category or unreadable; font determining means (3) that counts the number of unreadable characters in the font and determines the one with the least number of unreadable characters as the font for that page; and a dictionary of each font (
l_1, l_2, l_3) in sequence, and in the second step, the font dictionary (l_
1, l_2, l_3);
) and a threshold selection means (5) for selecting different judgment thresholds in the first stage and the second stage, the font is determined in the first stage for each page of the input document, and the font is determined in the first stage, A multi-font printed character reading device characterized in that, in the second stage, the dictionary of the determined font is checked and read using a judgment threshold different from that in the first stage.