JPS6327991A - Formation of histogram for input information recognizing device - Google Patents

Abstract

PURPOSE:To decrease the dimension number of the feature quantity of a dictionary and to increase the recognition speed of an input information recognizing device by integrating some of plural blocks divided according to the identification ability of the feature quantity, and thus generating a histogram. CONSTITUTION:An image information processor which uses an optical information input device consists of a keyboard 1 for KANA (Japanese- syllabary)/KANJI (Chinese-character) conversion or the like as well as character input, an input device 2 which reads an original optically, a processor 3 which performs operation in a mode specified by them, an output device 4 composed of a display device or printer, a program ROM 5 required for the operation of the device 3, a RAM 6 stored with common algorithm, and a word dictionary memory 7. Consequently, a direction code is added to the contour part of an input information pattern, which is divided into plural blocks according to the feature quantity identifying ability of the dictionary; and thus histogram of direction codes of each block is generated and compared and calculated by using the dictionary to determine a candidate and information according to the distance.

Description

[Detailed Description of the Invention] [Technical Field] The present invention is based on one character Ly! The present invention relates to a feature extraction technique for a character recognition device, and in particular to a histogram creation technique for a character recognition device.

[Prior Art] In an input device of an image information processing apparatus using a computer, for example, an optical information input device (hereinafter referred to as OCR) that optically reads and inputs information on a document, generally,
Input characters (input image information) are recognized using a recognition method based on a dictionary (template) matching method.

In character recognition devices such as OCR,
A direction code is attached to the outline of the input character pattern, the input character pattern is divided into multiple blocks, a histogram of the direction code is created for each divided block, and this histogram is compared with a dictionary prepared in advance. A candidate character is determined based on the calculated distance, and single character recognition is performed. That is, a histogram is created for each direction for the divided regions.

Using each histogram as a feature, distance calculation is performed,
In a character recognition device that determines characters, for example,
When the area is divided into 4×4, 128 (=4×4×8) dimensional features appear. Among these feature quantities, there are a mixture of those with high discrimination ability and those with low discrimination ability as a dictionary. Distance calculations are performed on these high and low values at the same level (handling).

However, in the above-mentioned method for creating a histogram in a character recognition device, the distance calculation is handled at the same level (handling) for features with high and low discrimination ability as a dictionary, so there is a problem that the recognition speed is slow. .

[Objective] An object of the present invention is to provide a technique that can increase the L& recognition speed of an input information recognition device.

Another object of the present invention is to provide a technique that can efficiently create a histogram for a single character recognition device.

The above and other objects and novel features of the present invention will become apparent from the description of this specification and the accompanying drawings.

[Configuration] In the present invention, a direction code is attached to the contour portion of an input information pattern, and the input information pattern is divided into a plurality of blocks according to the discriminating ability of the feature amount of a dictionary.

A histogram creation method for an input information recognition device that creates a histogram of the direction code for each divided block, compares this histogram with a dictionary prepared in advance, and determines candidate characters based on the distance, the method comprising: The present invention is characterized by comprising means for creating a histogram by integrating a portion of a plurality of blocks divided according to the discrimination ability of the feature amount of the dictionary.

[Example 1] An example of the present invention will be specifically described below with reference to the drawings.

In addition, in all the figures for explaining an example.

Components having the same function are given the same reference numerals, and repeated explanations thereof will be omitted.

FIG. 1 is a block diagram showing a schematic configuration of an OCR for carrying out a histogram creation method for an input information recognition device according to an embodiment of the present invention, and FIG. 2 shows image information using the OCR shown in FIG. FIG. 1 is a block diagram showing a schematic configuration of a processing device.

In FIG. 2, a keyboard 1 is used not only to input characters but also to designate various modes (kana-kanji conversion, kanji-kana conversion, OCR character recognition, etc.). The OCR input device 2 optically reads and inputs a document. The processing device 3 executes processing according to the specified mode on the input information from the keyboard 1 and OCR 2, and outputs it to the output device 4.

The output device 4 is a general term for a display device, a printer, etc. It is stored in the program memory (ROM) 5 that is necessary for the processing of the processing device 3, but it also performs kana-kanji conversion by keyboard input, post-processing of OCR character recognition, OC
R A common algorithm is used as much as possible for kana-kanji conversion and kanji-kana conversion of input characters. A data memory (RAM) 6 is used to store data and parameters that are being processed by the processing device 3. The word dictionary memory 7 stores a word dictionary to which reading orthography correspondence data is added.

As shown in FIG. 1, the OCR 2 shown in FIG. 2 reads and inputs image information such as characters on a document using an optical scanner 11 comprising a light source, a charge-coupled device (COD), and the like. Unit 1 extracts one character of image information such as the input kana character string, kana-kanji mixed character string, alphabetic character string, etc.
2, each character is extracted by the feature extraction unit 13.
Extract the features of the extracted characters. This extracted data is matched with feature dictionary data stored in a feature dictionary memory (ROM or RAM) 15 in a feature matching unit 14.

If matching is achieved, the input characters are Ly! , and sent to the processing device 3.

Next, FIG. 3 shows the process of dividing the input information pattern into a plurality of blocks (regions) according to the discriminating ability of the feature amount of the dictionary in the histogram creation method for an input information recognition device according to an embodiment of the present invention. The explanation will be given according to the flowchart shown below.

At step 101, a direction code (chain code) is added to the outline of the input information pattern, and the process moves to step 102. The direction-specific code is encoded using the direction vector shown in FIG.

In step 102, the number of codes attached to the outline of the input information pattern in each direction is counted, and the total number S is determined.Next, in step 103, the first dividing point (S/n) in the X direction is determined, and step At step 104, subsequent dividing points in the X direction are determined. For example, if the area is divided into n×m and the total number of codes is S, each division point is S / n + 2
S/n+...

(n-1)S/n coordinates, step 105
Find the coordinates of each division in the X direction.

Similarly, in step 106, the first dividing point in the Y direction (S/m)
In step 107, subsequent division points in the Y direction are determined. For example, S/m, 2S/m,..., (m-1)
The coordinates are S/m.

Next, in step 108, each division coordinate in the Y direction is determined, and the input image division processing is completed.

Next, the information pattern on the document is scanned in the X and Y directions, and the coordinates at which the number of codes corresponds to each dividing point are determined.

FIG. 5 shows the distribution of the histogram of each block of the dictionary when the area is divided into 4×4.

Comparing the 4 blocks in the center with the 12 blocks around it,
The number is quite small (i.e., the discrimination ability is low)
. Here, the larger the value, the greater the degree of dispersion (it is said that the greater the degree of dispersion, the greater the amount of information).

The most distinctive feature of the present invention is a method of creating a histogram by focusing on the central portion of the histogram where the degree of dispersion is small and integrating a plurality of blocks in this central portion.

Next, a process for creating a histogram in an embodiment of the present invention will be described with reference to the flowchart shown in FIG.

In step 201, the number of codes is counted in each direction for each block and a histogram is created.

At step 202, histograms in each direction are added for the central block, and the histogram creation process ends.

For example, in the input information pattern shown in Figure 7, a histogram is created for each of the 12 blocks around 1 (blocks excluding 6°7, 10, 11 in Figure 7), and the 4 blocks in the center are
Block (6, 7° 10.11) combines these into one to create a histogram.

Using the thus created 12x8+8=104-dimensional features and the dictionary, distance calculations are performed to select candidate information (for example, characters).

The present invention has been specifically explained above based on examples, but
It goes without saying that the present invention is not limited to the embodiments described above, and can be modified in various ways without departing from the spirit thereof.

〔Effect of the invention〕

As explained above, according to the present invention, by providing a means for creating a histogram by integrating a part of a plurality of blocks divided according to the discriminating ability of the feature amount of the dictionary, it is possible to Since the number of dimensions can be reduced, il
l speed can be improved.

Further, by reducing the number of dimensions of the feature amount, memory can be reduced.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing a schematic configuration of an OCR for implementing a histogram creation method for an input information recognition device according to an embodiment of the present invention, and FIG. 2 shows image information using the OCR shown in FIG. FIG. 3, a block diagram showing a schematic configuration of a processing device, shows an input information pattern in a histogram creation method for an input information recognition device according to an embodiment of the present invention, which is divided into a plurality of blocks (regions) according to the feature quantity discrimination ability of the dictionary. ) Flow chart of the processing process. FIG. 4 is a diagram showing the relationship between direction codes and direction vectors. FIG. 5 is a diagram showing the distribution of the histogram of each block of the dictionary when the area is divided into 4×4 areas. FIG. Flowchart of Process for Creating Histogram in One Embodiment FIG. 7 is a diagram showing an example of divided blocks of an input information pattern. In the figure, 3...processing device, 11...scanner, 12
. . . character cutting unit, 13 . . . feature extraction unit, 14 . . . feature matching unit, 15 . . . feature dictionary memory.

Claims (1)

[Claims] (1) Attach a direction code to the outline of the input information pattern, divide the input information pattern into multiple blocks according to the feature recognition ability of the dictionary, and create a histogram of the direction code for each divided block. A histogram creation method for an input information recognition device, in which candidate information is determined based on the distance by comparing the histogram with a dictionary prepared in advance, and determining candidate information based on the distance between the histogram and a dictionary prepared in advance. 1. A histogram creation method for an input information recognition device, comprising means for creating a histogram by integrating parts of a plurality of blocks.