JPH0962782A - Document reader - Google Patents

Abstract

PROBLEM TO BE SOLVED: To omit the image reduction processing, to reduce the image memory capacity, to shorten the processing time and to reduce the processing load by applying the area analysis to the image data inputted with low resolution, inputting only the area part in an image after the area analysis, and inputting the images in a mode corresponding to the area type after the area analysis. SOLUTION: A document reader consists of an image input part 1, an image memory 2, an area analysis part 3, an area memory 4, a character recognition part 5, etc. In such a constitution, the characters of images recorded on a recording medium are scanned for production of the image data of low resolution. Then the area data are generated based on the image data. The characters or images recorded on the recording medium are scanned again based on the area data, and the image data are produced for every area. Furthermore, the image data on the character areas are recognized based on the area data, and the recognition data are produced. Then the optional document data are produced from the image data, the area data and the recognition data and outputted to an output medium.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a document reading device for reading characters and images (graphics other than characters, pictures, photographs, ruled lines, etc.) recorded on a document.

[0002]

2. Description of the Related Art FIG. 7 is a flow chart of a reading process in a conventional document reading apparatus, particularly a process of scanning with high resolution from the beginning without prescanning. The reading process in the conventional document reading device will be described with reference to FIG. First, an image of a document to be read is input with high resolution (S101). The high-resolution image input in S101 is displayed on the display device, and the operator specifies a range necessary for reading with a frame (S103). If the reading range is the entire page, this frame designation process is unnecessary.

Next, a reduced image required for area analysis is created (S105). FIG. 8 is an explanatory diagram illustrating the creation of a reduced image. In FIG. 8, 210 is the high resolution image data input in the processing of S101, and 211 is the S resolution.
It is the reduced image data created in the process of 105. In this reduction processing, for example, if the high resolution image of 8 × 8 dots has one or more black pixels, the reduced image 1 dot is black, and if the high resolution image is all white pixels, the reduced image 1 dot is white. That is, high resolution image data is not required for area analysis, and rough (low resolution) image data is sufficient. Next, a region is extracted from the reduced image created in S105, and a region analysis is performed for this region to distinguish characters and images from image features (S10).
7). Then, the character of the high resolution image data is recognized based on the result of the area analysis (S109). further,
The recognition result is output by storing the recognition result in the memory or printing the result on the printer (S1).
11).

FIG. 9 is a flowchart of a reading process in another conventional document reading apparatus, in which a reading frame is designated by prescanning at a low resolution. FIG.
Based on the above, the reading process in the conventional document reading device will be outlined. First, a low-resolution image is pre-scanned to input a low-resolution image (S121), the input low-resolution image is displayed on a display device, and the operator designates a reading range with a frame (S123). Next, the reading frame portion is scanned to input an image with high resolution (S125). After this, the same processing as S105 to S111 shown in FIG. 7 is performed.

[0005]

However, the conventional method described above has the following problems. That is,
It is necessary to perform processing for creating a reduced image, which is the processing in S105. Further, rescanning is necessary in order to obtain an image (multivalue, color, etc.) other than the high resolution image necessary for the processing after the recognition processing (S109), but when rescanning after area analysis, It takes a long time to rescan. Furthermore, in the processing flow shown in FIG. 7, a high-resolution image other than the necessary area is input, which wastes time.

[0006]

A document reading apparatus according to the present invention scans a character or an image recorded on a recording medium, converts it into an image signal, and digitally converts this image signal to obtain low-resolution image data. A low-resolution image input means, a low-resolution image memory that stores this image data, an area analysis means that creates area data based on the low-resolution image data in this low-resolution image memory, and this area data An area memory for storing, area image input means for rescanning a character or an image recorded on the recording medium according to area data in the area memory to create necessary image data for each area, and this image According to the area image memory for storing the data and the area data in the area memory, the image data of the character area in the area image memory is recognized. Character recognition means for creating identification data, a recognition memory for storing this recognition data, image data in the area image memory, area data in the area memory, and arbitrary document data from the recognition data in the recognition memory And a result output procedure for outputting this document data to an output medium.

In the document reading apparatus configured as described above, a character or an image recorded on a recording medium is scanned to create low resolution image data, and area data is created based on the low resolution image data. To do. And
The characters or images recorded on the recording medium are rescanned in accordance with the area data to create the necessary image data for each area. Further, according to the area data, the image data of the character area is recognized and recognition data is created. Furthermore, arbitrary document data is created from image data, area data, and recognition data, and this document data is output to an output medium.

[0008]

BEST MODE FOR CARRYING OUT THE INVENTION

Embodiment 1. FIG. 1 is a block diagram showing a configuration of the first exemplary embodiment of the present invention. In the figure, reference numeral 1 denotes an image input unit, which optically scans all / a part of a document to be read, converts characters and images recorded on the document into an image signal by photoelectric conversion, and further, this image The signal is converted into binary / multi-value / pseudo multi-value / color image data. Two
Is an image memory, which stores image data output from the surface image input unit 1. An area analysis unit 3 extracts an area from (low resolution binary) image data in the image memory 2,
Region data is created by identifying each region as a character and an image. An area memory 4 stores area data output from the area analysis unit 3.

A character recognition unit 5 cuts out a character image of this image data for each character from the area data in the area memory 4 and the image data in the image memory 2, recognizes the character, and converts it into a character code. To do. Then, this character code is used as recognition data. Further, the character recognition unit 5 can also perform knowledge processing on the recognition data and correct it. A recognition memory 6 stores the recognition data output from the character recognition unit 5. A result output unit 7 includes image data in the image memory 2, area data in the area memory 4, recognition data in the recognition memory 6 (image data, area data,
Document data is created from at least one of the recognition data), and this document data is printed by the printer 8 or stored in the output memory 9. Then, via this output memory 9, another document data processing system (word processor, DTP
The document data is passed to a system, a document management system, etc. (not shown). Alternatively, the document data can be passed to another document data processing system by communication (not shown).

Reference numeral 10 is a display unit composed of a CRT or the like.
Reference numeral 11 denotes an operation unit including a keyboard and a mouse,
The interface between the operator and the document reading device is used to start / end the reading process (whole or part), display the processing result (image data, area data, and recognition data) and confirm / correct the process. is there.

FIG. 2 is a flowchart of the operation of the embodiment configured as described above. The operation of this embodiment will be described below with reference to FIG. First, the image input unit 1 performs a prescan for inputting an image at a low resolution (S
1). In the prescan, the image input unit 1 scans an input document to be read, converts characters and images recorded on the document into an image signal by photoelectric conversion, and further converts the image signal into a binary low resolution (for example, 50 DPI). ) Image data. Then, this image data is stored in the image memory 2.

When the image input is completed, the operator uses the display unit 10 and the operation unit 11 to specify the reading frame (S).
2). That is, the image data in the image memory 2 is displayed on the display unit 10, and the operator views the display and designates a portion required for reading as a reading frame (rectangle, polygon) using the operation unit 11. The reading frame designation is used when a part of the image data is read, such as avoiding a shadow caused by copying or reading only one article of a newspaper. Also, in order to simplify the operation,
If no reading frame is specified, all image data will be read.

When the reading frame designation is completed, the area analysis unit 3
The area is analyzed by (S3). In the area analysis, first, the area analysis unit 3 uses the marginal distribution histogram of black pixels or characters and images from the image data in the image memory 2 corresponding to the reading frame, or the run length method. Use to extract the region frame. And
Character regions and images (figure / figure / figure / figure / figure / figure)
(Photo) area. Then, this area frame and its type are used as area data.

This area analysis method is the same as the area analysis for the reduced image. The area data and the image data are overlapped and displayed on the display unit 10, and the operator confirms / corrects the area data. Area data after confirmation / correction is stored in area memory 4
To be stored. FIG. 3 shows an example of the image data (the reading frame is not specified: the reading target is all the image data) in the area data. The image data is subjected to area analysis, and a Japanese area 21, an English area 22, a graphic area 23 and a photograph area 2
Extract 4. Table 1 shows this area data in a table.

[0015]

[Table 1]

In Table 1, "Xs, Ys" is the coordinates of the upper left apex of the area, and "Xe, Ye" is the coordinates of the lower right apex of the area (the area is represented by a rectangle, and the resolution is 50 DPI during prescan). (Represented by the number of dots in), "Type" is the identification result of the area. When the area analysis is completed, the image input section 1 performs image input scanning in a mode corresponding to each area (S4). In this scan, the image input unit 1 inputs the image of the portion corresponding to the area data in the area memory 4 in the corresponding image input mode.

FIG. 4 is a flow chart of an example of this scan, and this scan will be described below with reference to FIG.
First, rectangles (39, 59) and (374, 492) including the entire area are calculated from the area data in the area memory 4 (S).
21). Table 2 is a table showing the types of objects to be read and the image input modes corresponding thereto.

[0018]

[Table 2]

As shown in Table 2, the maximum resolution in the image input mode of the image data shown in FIG. 3 is 400 for a Japanese sentence.
It is DPI, and when the above rectangle is converted into 400 DPI, (312, 472) becomes (2992, 3936) (S22). This rectangle (312, 472), (299
2, 3936) with a maximum resolution of 400 DPI and multi-valued image input (S23).

Table 3 shows the vertices of each area shown in Table 1 converted at the maximum resolution of 400 DPI.

[Table 3]

According to Table 3, the multi-valued image data input in S22 is extracted for every four areas (S24). The extracted image data is converted for each area according to the resolution and conversion method shown in Table 2 (S25). The converted image data is stored in the image memory 2 for each area (S26), and the scanning process ends. When the scan (S4) changes, the character recognition unit 5 recognizes the character (S5). The outline of this character recognition method is as follows. First, a line is cut out from the image data in the image memory 2 corresponding to the character area of the area data in the area memory 4 by using the peripheral distribution histogram of black pixels or the run length method. In addition, cut out characters from the line. Next, the cut-out character image is subjected to a character code (including a candidate character and a certainty factor of the candidate character) by pattern matching using a recognition dictionary that stores the standard character recognition features in the character recognition unit 5.
Convert to Further, the character code can be knowledge-processed using a word dictionary in the character recognition unit 5 or a knowledge dictionary storing grammatical rules to correct the character code (compliance of candidate words and candidate words is created). it can.

The character code (including candidate characters and candidate words) and the certainty factor are used as recognition data. This recognition data is displayed on the display unit 10 (compared with the image in the character area), and the operator confirms / corrects the recognition data. The recognition data after confirmation / correction is stored in the recognition memory 6. When recognition is over,
The result output unit 7 outputs the reading result (S6).
That is, the result output unit 7 includes the image data in the image memory 2, the area data in the area memory 4, and the recognition data in the recognition memory 6 (at least one of the image data, the area data, and the recognition data). Document data is created from the data, and this document data is printed by the printer 8 or stored in the output memory 9.

As described above, according to the first embodiment, the following effects can be obtained. (1) Since image reduction is unnecessary by performing area analysis on image data input at low resolution, it is possible to reduce processing time and the processing load of the document reading apparatus. (2) Since only the area is input as an image after area analysis,
It is possible to reduce the image memory capacity because the image data of the blank portion other than the area is not necessary, and it is possible to shorten the processing time after the scan. (3) Since the image is input in the mode corresponding to the area type after the area analysis, image data suitable for the processing after the recognition can be obtained, and the processing steps can be reduced (for example, the image input by the pseudo multi-valued image of the photograph portion can be performed). Unnecessary), and processing time can be shortened. (4) When the operator finds an image error (for example, inputting a document not to be read, wrong paper orientation, excessive skew, etc.) in the processing from prescan to area analysis, and represcan. The processing time can be shortened as compared with the conventional method.

Embodiment 2. In this embodiment, the "scan" process (S4) in the first embodiment is performed by a different method, and the other processes are the same as those in the first embodiment. FIG. 5 is a "scan" according to the second embodiment of the present invention.
It is a flowchart of FIG. Hereinafter, the second embodiment will be described with reference to FIG. First, from the area data in the area memory 4, rectangles (39, 59), (374, 49) including the entire area are included.
2) is calculated (S31). This processing is the same as in the first embodiment. From Table 2 and the area data, the maximum resolution in the image input mode is 400 DPI, and the rectangle is 400 DP
Convert with I. This converted value is (312, 472), (2
992, 3936). Then, a rectangle that is slightly larger than the rectangle indicated by these coordinates, for example, rectangles (232, 392), (3 that are larger by about 5 mm vertically and horizontally).
072, 4016) is set (S32).

An image of this rectangle is input with a maximum resolution of 400 DPI and multiple values (S33). This processing is the same as in the first embodiment. According to Table 3, the multi-valued image data input in S33 is corrected for the deviation in each of four areas to extract the image data (S34). FIG. 6 is an explanatory diagram illustrating a method of correcting the deviation. A method of correcting the deviation will be described with reference to FIG. Data is checked for each side (scanning line) of the rectangle formed by the coordinates shown in Table 3, and if there is any non-white pixel (non-zero) on each side, scanning is performed in the direction to expand the frame. The line is moved, and the scanning line is moved until the data on each side is only white pixels. On the other hand, when only white pixels are present on each side, the scanning line is moved in the direction of narrowing the frame, and the scanning line is moved until a non-white pixel appears on the scanning line.

The rectangular data in the area memory 4 is updated to the one in which the deviation is corrected by the process of S34 (S35). The extracted image data is converted for each area according to the resolution and conversion method shown in Table 2 (S36). This processing is the same as in the first embodiment. Next, the converted image data is stored in the image memory 2 for each area (S37). This process is also the first embodiment
Is the same as

As described above, according to the second embodiment, the deviation of the image data between the prescan and the scan (the cause of the deviation: the difference in resolution, the movement of the input document, the temporal change of the scanner unit, etc.) is corrected. Since the image data of the area can be accurately extracted, the lack of image data and blank spaces can be avoided. Therefore, the reading process can be performed accurately.

In the above-described first and second embodiments, an example in which multi-value scanning is performed has been described, but it is also possible to scan in the image input mode (color, binary) required for result output. It is also possible to scan in a designated image input mode for each area. In this case, this function is required in the scanner section. Further, in the second embodiment, the multivalued image data is processed by the deviation correction in the processing of S34. However, once the multivalue is converted into binary, the deviation correction is performed on the binary image data. You can also do

[0029]

As described above, according to the present invention, the following effects can be obtained. By performing area analysis of image data input at low resolution, image reduction is not necessary, so that it is possible to reduce the processing time and the processing load of the document reading device. Further, since only the area part is image-inputted after the area analysis, the image data of the blank part other than the area is unnecessary, the image memory capacity can be reduced, and the processing time after scanning can be shortened. Furthermore, since the image is input in the mode corresponding to the area type after the area analysis, image data suitable for the processing after recognition can be obtained, and the processing steps and the processing time can be reduced.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of the present invention.

FIG. 2 is a processing flowchart of an embodiment of the present invention.

FIG. 3 is a diagram showing an example of image data according to the embodiment of the present invention.

FIG. 4 is a flowchart of scan processing according to the first embodiment of the present invention.

FIG. 5 is a flowchart of scan processing according to the second embodiment of the present invention.

FIG. 6 is an explanatory diagram illustrating a shift correction according to the second embodiment of the present invention.

FIG. 7 is a flowchart of a reading process in a conventional document reading device.

FIG. 8 is an explanatory diagram illustrating generation of a reduced image in a conventional document reading device.

FIG. 9 is a flowchart of a reading process in another conventional document reading device.

[Explanation of symbols]

 1 image input unit 2 image memory 3 region analysis unit 4 region memory 5 character recognition unit 6 recognition memory 7 result output unit 8 printer 9 output memory 10 display unit 11 operation unit 12 control unit

Claims (2)

[Claims] 1. A low resolution image input means for scanning a character or an image recorded on a recording medium to convert the image signal into an image signal, and digitally converting the image signal to generate low resolution image data, and the image data. A low-resolution image memory, area analysis means for creating area data based on the low-resolution image data in the low-resolution image memory, an area memory for storing the area data, and area data in the area memory. Area image input means for rescanning the characters or images recorded on the recording medium according to the above to create necessary image data for each area, an area image memory for storing the image data, and an area image memory Character recognition means for recognizing the image data of the character area in the area image memory and creating recognition data according to the area data of A recognition memory for storing recognition data, image data in the area image memory, area data in the area memory, and arbitrary document data is created from the recognition data in the recognition memory, and the document data is used as an output medium. A document reading device comprising: a result output manual output. 2. The document reading apparatus according to claim 1, wherein the area image input means has a function of correcting a deviation between the area data in the area memory and the rescanned image data.