JPH08221515A - Image processor - Google Patents

Abstract

PURPOSE: To improve the character recognition rate by highly accurately and stably binarizing the character area. CONSTITUTION: An area identifying part 104 roughly identifies the character/ table area and the other areas on an input document. A threshold value control part 120 decides the binarization threshold value based on a density histogram at the entire part or one part of the character/table area. While using the binarization threshold value set by the threshold value control part 120, a multilevel/binary converting part 108 converts the multilevel expression image data in the character/table area on the input document to a binary expression by performing a simply binarizing processing. The binarization threshold value is fixed in the same area, at least. The multilevel expression image data in the other areas are converted into the binary expressions by bither processing by a multilevel/binary converting part 110.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus for handling multi-valued image data, and more particularly to an image processing apparatus for converting multi-valued image data into binary image data.

[0002]

2. Description of the Related Art In recent years, the field of handling multi-valued image data obtained by scanning a document such as a document in which areas having different attributes such as characters and photographs are mixed is expanding. Particularly in various fields such as image recognition, character recognition, image communication, image storage, etc., a process of converting such multivalued image data into binary image data is often required, and a processing method therefor is required. Many have been proposed.

For example, for each small block such as 4 × 4 pixels or 3 × 3 pixels, the attribute determination is performed to determine whether the image is a binary image or a character image, or a grayscale image or a pattern. A processing method is known in which simple binarization processing with a single threshold is applied to blocks that are determined to be present, and dither processing is applied to blocks that are determined to be grayscale images or patterns (Shinji Tetsuya et al. "Binarization processing method for manuscripts in which binary images and grayscale images coexist", SIJ Vol.J67-BNo.7, p.
p.781-788 and Satoshi Ouchi et al., "Character / Dot /
Image Area Separation Method for Mixed Photo Image ", IEICE Technical Report IE90-3
2).

Further, a method of detecting a ridge point and a valley point, which are structural information of a multivalued image, and adjusting a binarization threshold value for each scanning line so that a binarization error of the ridge point and the valley point is reduced. Is known (Satoshi Ouchi et al., "Input device with document image preprocessing by ridge point / valley method", IEICE Tech.
5).

[0005]

However, in the above-mentioned method of switching the binarization processing method in small block units,
The inconvenience that the binarization processing method is switched within the area of the same attribute on the document often occurs. When the image data after the binarization processing is used for character recognition, if an improper binarization processing method is selected within the character area on the original, it may cause erroneous recognition at that portion. It is a very serious problem because it is directly related to the occurrence.

In the above-mentioned method of adaptively controlling the binarization threshold value for each scanning line, the binarization threshold value fluctuates within the same region, and therefore the same problem tends to occur. further,
Since the region is not taken into consideration when determining the binarization threshold value, there is a problem that the character region cannot be binarized with the optimum threshold value when the character region, the photo region, and the like are mixed.

The present invention is intended to provide an image processing apparatus which solves the above-mentioned problems relating to the conversion processing of multi-valued expression image data into binary expression, and in particular, a pre-processing apparatus for a character recognition apparatus. It is an object of the present invention to provide an image processing device that is suitable for use as a computer and can contribute to an improvement in the character recognition rate.

[0008]

In order to achieve the above-mentioned object, the image processing apparatus according to the invention described in claims 1 to 5 has the following features.

According to a first aspect of the present invention, there is provided area identifying means for globally identifying an area in which characters are written on the original and other areas from multivalued image data of the original.
The multi-valued representation image data of the area in which the character identified by the area identification means is written is converted into 2 by a certain processing method.
And multi-value / binary conversion means for converting the image data into the value-represented image data.

According to a second aspect of the present invention, in the configuration of the first aspect of the invention, the multi-value / two-value conversion means performs multi-value representation image data by simple binarization processing with a constant binarization threshold value. Is a means for converting the image data into binary representation image data.

According to a third aspect of the present invention, there is provided area identifying means for globally identifying an area in which characters are written on the original and other areas from multivalued image data of the original.
The multi-valued binary conversion means for converting the multi-valued expression image data of the area in which the character identified by the area identification means is written into the binary expression image data by the simple binarization processing, and the area identification means. A density histogram is obtained from the multivalued image data of the region in which the identified character is written, and means for determining a binarization threshold value for the multivalue-binary conversion means is provided based on the density histogram. However, the binarization threshold is fixed at least in the same region.

According to a fourth aspect of the present invention, there is provided area identifying means for globally identifying an area in which characters are written on the original and other areas from multivalued image data of the original.
The multi-valued representation image data of the area in which the character identified by the area identification means is written is converted into 2 by simple binarization processing.
A multivalue-binary conversion means for converting into value expression image data, and a part of the area in which the character identified by the area identification means is written is selected, and the multivalued representation of the selected area A density histogram is obtained from the image data, and based on the density histogram, 2 for the multi-value / binary conversion means is obtained.
Means for determining a binarization threshold value, and the binarization threshold value is fixed at least in the same region.

According to a fifth aspect of the present invention, in the structure of the third or fourth aspect of the invention, the means for determining the threshold value obtains the density histogram by excluding the solid white or black solid portion inside the area region. Is.

[0014]

According to the first aspect of the present invention, the area on the manuscript in which the characters are written (as will be described in detail in connection with the embodiments described later, "character area" including only characters and ruled lines ") A table area ") is globally identified, and the multivalued representation image data is converted into a binary representation by a certain processing method. According to the second aspect of the present invention, the multi-valued representation image data of the area in which the character is written is converted into the binary representation image data by a certain simple binarization process. Therefore, unlike the conventional local method or the method of controlling the binarization threshold value for each scanning line, the processing method does not change or the binarization threshold value does not change within the area where the character is written. Therefore, stable binary representation image data can be obtained.

According to the third aspect of the invention, the multivalued representation / binary representation conversion of the area in which the character is written is performed by a simple binarization process, and the binarization threshold value is the character written. It is optimized based on the density histogram of the area, and the binarization threshold is fixed at least in the same area. Therefore, it is possible to perform highly accurate and stable binarization on a region in which characters are written. Regarding the correspondence between the binarization threshold value and the area, as will be described in connection with the embodiment described later, a mode in which the binarization threshold value is determined based on the density histogram of each area, and the density of all the area areas A mode in which the binarization threshold value for the entire region is determined based on the histogram is included.

According to the fourth aspect of the present invention, in order to determine the binarization threshold value, only the density histogram of a partial area selected from the area in which the character is written is used. The amount of processing for determining the activation threshold can be small, and the processing can be speeded up. Further, as will be described in connection with the embodiment described later, if one area or a predetermined number of areas estimated to have a large number of characters is selected, the number of areas to be used is at least 2.
It is possible to optimize the binarization threshold sufficiently and perform highly accurate binarization.

If a large white or black solid area exists inside the area in which the characters are written, the density histogram may deviate from the original distribution. However, according to the fifth aspect of the invention, the density histogram is obtained by excluding the white or black solid portion inside the area where the character is written, and the binarization threshold value is determined based on the density histogram. Even if there are white or black solid areas in the area,
Highly accurate binarization is possible with the optimal binarization threshold.

[0018]

Embodiments of the present invention will be described below with reference to the accompanying drawings. <Embodiment 1> FIG. 1 is a schematic block diagram of an image processing apparatus according to this embodiment. In FIG. 1, image data 100 of a multivalued representation of a document such as a document is read from a scanner (not shown) or the like.
Is input, and the multivalued representation image data 100 is stored in the image memory 102.

The area identification unit 104 reads the multi-valued expression image data in the image memory 102, identifies each area having different attributes on the document, and cuts it out as a rectangular area. Figure 2
It is explanatory drawing of this area | region identification. When the multi-valued representation image data of the illustrated document 200 is input, the character areas 202 to 205, the table area 206, and other areas (enclosing frame 207, horizontal separator 208, vertical separator 20) on the document are input.
9, FIG. 210, and further, a photograph and a halftone dot image area) are identified and each is cut out as a rectangular area.

As can be understood from this example, the area identification unit 104 performs global area identification, that is, a rectangular area surrounding a portion (column) of a character line or a plurality of character matrices on a document, and vertical and horizontal ruled lines. It identifies and cuts out a region having a certain size, such as a rectangular region surrounding a table consisting of, and does not identify the attribute in units of small blocks as in the prior art. In addition,
The character area in which the characters are written and the table area are areas for character recognition, but when it is not necessary to distinguish them, the character area and the table area may be cut out as areas having the same attribute.

Such global area identification can be performed by various methods, and for example, the method of the applicant's Japanese Patent Application No. 5-159190 or Japanese Patent Application No. 5-331252 can be used. The details of the methods of both patent applications should be referred to the specification and drawings, and the outline thereof is as follows. First, since the input image data has a multi-valued representation, as preparation, it is converted into binary representation image data by simple binarization with an appropriate threshold value. Next, the binary-represented image data is compressed with an appropriate magnification, a rectangle circumscribing the black pixel connected component is extracted from the obtained compressed image data, and the obtained rectangle is classified into a character rectangle and other rectangles. . And
A character area is extracted by integrating the character rectangles. The other rectangles are also classified and integrated as necessary to extract the area. The image memory 102 may be used as a storage area for the binary representation image data and the compressed image data during this processing.

Further, among the other areas, the area of the halftone dot image or the photograph may be positively identified. For example, similar to the above-mentioned “image area separation method of mixed image of characters / dots / photos”, determination of dots / photos on a block-by-block basis and integration of dots or blocks of photo to obtain respective areas Can be extracted.

The result of such area identification is passed to the control section 106 which controls each section of the apparatus. When the area identification processing is completed, the control unit 106 sequentially reads the multi-valued expression image data from the image memory 102 and outputs three types of multi-valued / two-value conversion units 1.
08, 109, 110. At this time, one of the three types of multi-value / binary conversion units 108, 109, 110 is selected and validated according to the type of the region to which the data belongs. That is, when the pixel data in the character area is read, the multi-value / binary conversion unit 108 is read, and when the pixel data in the table area is read, the multi-value / binary conversion unit 109 is used, and the other area is read. When the data of the pixel inside is read out, it is multi-valued-2
The value converters 110 are enabled respectively. The three types of multi-value / binary conversion units 108, 109, and 110 respectively convert multi-valued expressions into binary expressions. Value-to-binary converter 1
Reference numerals 09 respectively convert the multivalued representation image data into the binary representation image data by a simple binarization process using one threshold value designated in advance, and the multivalued-to-binary conversion unit 11 for other areas.
A value of 0 performs conversion from multi-valued representation to binary representation by, for example, dither processing.

When the other areas are classified into a figure / separator area and a halftone / photo area, the multi-value / binary conversion unit 110 converts the data of the halftone / photo area by dithering, -Simple data for separator area 2
It can also be converted by a digitization process. Further, depending on the use of this device, the multi-valued representation image data of other areas may be output as it is.

As described above, since the respective areas on the original cut out by the global area identification are converted into the binary representation by the same processing method, the same areas as in the conventional local method are used. It is possible to obtain stable binary representation image data for the character area and the table area without switching the processing method inside. Therefore, this device is used as a pre-processing device for a character recognition device and multivalued-2
By inputting the binary representation image data obtained by the value conversion units 108 and 109 to the character recognition device, it is possible to reduce the erroneous recognition due to the binarization failure and expect to improve the recognition rate.

<Second Embodiment> FIG. 3 is a schematic block diagram of an image processing apparatus according to the present embodiment. In FIG. 3, the same reference numerals as those in FIG. 1 indicate the same parts or corresponding parts, and the description thereof will be omitted. The difference from the first embodiment shown in FIG.
2 of the data of the character area and the table area that are the target of character recognition
The common multi-value / two-value conversion unit 111 performs the conversion into the value expression by a simple binarization process, and the binarization threshold value used in the multi-value / two-value conversion unit 111 is threshold-controlled. The point is that the control is performed by the unit 120.

This device has the following modes for controlling the binarization threshold of character / table area data, and the user can arbitrarily specify one mode.

Mode: For each area of the character and the table, a threshold value is obtained based on the density histogram of the area. In this mode, when obtaining the density histogram of the character area, you can specify whether to exclude the white or black solid area in the area, but this sets the threshold value using the density histogram of the character area. The same applies to other modes.

Mode: A threshold is obtained based on the density histograms of all character areas, and this threshold is applied to all character areas and table areas.

Mode: One character area is selected, a threshold value is obtained based on the density histogram of this character area, and this threshold value is applied to all character areas and table areas. In this mode, as a character area selection method, a method of selecting a character area with the largest area, a method of selecting a character area with the largest number of lines, and a method of selecting a character area with the largest number of characters can be specified.

Mode: A threshold value is obtained based on the density histograms of all table areas, and this threshold value is applied to all character areas and table areas.

Mode: Select the table area with the largest area,
A threshold value is obtained based on the density histogram, and this threshold value is applied to all character areas and table areas.

FIG. 4 is a block diagram of the threshold controller 120. In FIG. 4, 122 is a solid area removing unit for removing solid white or black areas in a character area, 123 is a histogram creating section for obtaining a density histogram of a character area or a surface area, and 124 is based on the created density histogram. A threshold value calculation unit that calculates a threshold value and a threshold value setting unit 125 that sets the calculated threshold value to the multi-value to binary conversion unit 111. The threshold control operation in each mode will be described below.

First, the case where the mode is designated will be described. The control unit 106 sequentially selects the character region and the table region identified by the region identification unit 104 one by one, reads the multivalued representation image data from the image memory 102, and inputs the multivalued representation image data to the threshold value control unit 120.

The image data input to the threshold control unit 120 passes through the solid area removing unit 122 and the histogram creating unit 1
However, if the removal of the white or black solid portion in the character area is designated, the solid portion removal unit 122 receives the image data of the character area, and the solid portion removal unit 122 has a certain area or more solid white or black area. Only the image data excluding the copy is sent to the histogram creation unit 123.

In the example of the character area 230 of FIG. 5, since the rectangular range enclosing the character lines 231 to 235 having different line lengths is cut out as one area, a fairly wide white solid portion 236, in which no character exists, is formed. 237 is included. Since noise due to show-through etc. is likely to be introduced into such white solid areas, if a density histogram is obtained including a large white area, a peak appears in the density histogram at an extremely thin density value. It is likely to be adversely affected when determining the threshold value based on. Further, even when a wide black solid portion is left, the deviation from the original density histogram becomes large, and the threshold value determination is likely to be adversely affected. Therefore, a white or black solid part having a certain area or more is removed.

It should be noted that the same effect can be obtained by subdividing the character area in the area identifying unit 104 so that the character area does not include a solid white or black area having a certain area or more. , Which is also included in the embodiments of the present invention. In the example of the character area 230 shown in FIG. 5, the area surrounding the character lines 231 to 233 and the character lines 234 and 235.
The white solid part 23 having a large area
6,237 are removed.

The histogram creating section 123 creates a density histogram of image data of the character or table area input via the solid area removing section 122. The threshold calculator 124 calculates a threshold for binarization based on this density histogram.

Generally, in the character area (or table area),
Although a bimodal density histogram as shown in FIG. 6A is obtained, a density histogram as shown in FIG. 6B is obtained in a photographic region or the like. Even if the same binarization method or threshold value is applied to the regions having completely different density distributions as described above, good results cannot be obtained. Therefore, for the purpose of improving the character recognition performance, the binarization threshold value for the character area or the table area is determined based on the density distribution of the area.

Now, basically, it is sufficient to select the density value near the valley of the density histogram as shown in FIG. 6A as the binarization threshold value, but the density histogram actually obtained is not necessarily shown in FIG. Often, the valleys are not as clear as shown in (). Therefore, conventionally, various methods have been devised for obtaining the optimum binarization threshold value based on the density histogram. The threshold value calculation unit 124 may adopt such a conventional appropriate threshold value determining method or another method. Here, paying attention to the fact that the amount of calculation is small, a known method of determining the threshold value by using the 0th and 1st order cumulative moments of the density histogram (Otsuno Nobuyuki, “discrimination and an automatic method based on the least squares criterion”). Threshold value selection method ", Theological theory '80 / 4 Vol.J63-D No.4)
The threshold shall be determined. The details of the process are detailed in the same paper, so the description is incorporated by reference.

The threshold value thus determined is stored in the threshold value setting section 125 in association with the area. When the thresholds for all the character areas and the table areas are determined, the control unit 106 sequentially reads out the image data in the image memory 102 and sends the image data to the multi-value / binary conversion units 110 and 111. The multi-value to binary conversion unit 111 is enabled when the data is in the area of the table, and the multi-value to binary conversion unit 110 is enabled when the data is in the other area. Further, when the data of the character area or the table area is read, the identification information of the area is given to the threshold control unit 120. The threshold value setting unit 125 of the threshold value control unit 120 determines which character or table area data is being processed based on this identification information, and multi-values the threshold value (stored inside) corresponding to the area. It is set in the -2 value conversion unit 111.

As described above, according to this mode, the threshold value for binarization is adaptively controlled according to the density histogram of each area of each character or table, and the threshold value is set within each area. Is constant, the image data of the character area and the table area is highly accurate and stable even when the density distribution varies depending on the area.
It can be valued. Therefore, the multi-value / binary conversion unit 11
If the output data of 1 is input to the character recognition device, a significant improvement in the character recognition rate can be expected.

Next, the case where the mode is designated will be described. The control unit 106 sequentially selects the character regions identified by the region identification unit 104 one by one, reads the multi-valued representation image data from the image memory 102, and inputs it to the threshold value control unit 120. The image data input to the threshold value control unit 120 is sent to the histogram creation unit 123 via the solid portion removal unit 122. When the removal of the white or black solid portion in the character area is designated, when the image data of the character area is input, the solid portion removing unit 122 removes the image of the white or black solid portion having a certain area or more. Only the data is sent to the histogram creation unit 123. The histogram creation unit 124 creates a density histogram from the input image data without distinguishing the area. When the image data of all character areas is input and the density histogram is obtained, the threshold calculation unit 125 obtains the threshold value by the above-mentioned method based on this density histogram. This threshold is stored inside the threshold setting unit 125.

When the threshold value is thus determined, the control unit 106 sequentially reads the image data in the image memory 102 and sends it to the multi-value / binary conversion units 110 and 111, and the image data is a character or a display. The multi-value to binary conversion unit 111 is enabled for the data of the area, and the multi-value to binary conversion unit 110 is enabled for the data of the other area. The threshold value setting unit 125 of the threshold value control unit 120 sets the stored threshold value in the multi-value / binary conversion unit 111.

As described above, according to this mode, the threshold value for binarizing each area of the character and the table is adaptively controlled according to the density histogram of the entire character area, and the threshold value is set within each area. Is constant. Therefore, it is possible to perform accurate and stable binarization of the character area and the table area.
However, since the threshold value is controlled based on the density distribution averaged over all the character areas, the binarization accuracy becomes worse than that of the mode when the density distribution greatly varies depending on the area.

It should be noted that instead of obtaining the density histograms for all the character areas, a predetermined fixed number of character areas are selected, for example, from the beginning of the document, and the threshold value is determined based on the density histogram. Of course, this is also included in the embodiments of the present invention.

The case where the mode is designated will be described. The control unit 106 acquires necessary information from the region identification unit 104 and selects one character region for creating a density histogram. When the selection of the character area with the largest area is designated, one area is selected by grasping the area from the position information of each character area acquired from the area identification unit 104 and comparing them. When the selection of the character area with the maximum number of lines is designated, the area identification unit 104
The number of character lines contained in each character area is obtained, and this selects the largest character area. When the selection of the character area of the maximum number of characters is designated, the number of characters in each character area is estimated from the number of rectangles and the character size in each character area obtained during area identification in the area identification unit 104, Select the largest character area. In other words, we are trying to select a character area with as many characters as possible, but the processing load for this selection is naturally the lightest using the area, and the lightest using the number of lines, The method of directly estimating the number of characters is the heaviest.

When one character area is selected in this way, the image data of this character area is read from the image memory 102 and sent to the threshold value controller 120. Threshold control unit 120
In the input image data, the solid image removal unit 12
2 is sent to the histogram creation unit 123 as it is,
Alternatively, when the solid portion removal is designated, the data of the white or black solid portion of the character area is removed and sent to the histogram creation unit 123. Then, the histogram creation unit 123 creates a density histogram, the threshold calculation unit 124 determines a threshold value based on the density histogram, and the threshold value is stored in the threshold value setting unit 125 and set in the multi-value / binary conversion unit 111. .

Next, the control unit 106 sequentially reads out the image data from the image memory 102, and the multi-value / binary conversion unit 110,
Although the data is sent to 111, the multi-value to binary conversion unit 111 is enabled when reading the data of the character or table area, and the multi-value to binary conversion unit 110 is enabled when reading the data of the other area.

In this mode, since the threshold value is determined by using only one character area, the processing amount for determining the threshold value is small and the corner processing can be speeded up. Moreover, since the character area estimated to have the largest number of characters is selected as a representative, it is possible to determine a threshold value that improves the total recognition rate. That is, the character area selected as a representative for determining the threshold value is usually one of the body areas. The text area other than the text, such as headings, is often small in size, and the text size is often larger than the text. In addition, it is often written in a special typeface such as Gothic. When binarization is performed using the threshold value determined based on the density histogram of, the recognition rate of headings and the like may decrease. However, since the ratio of the number of characters such as a headline to the total number of characters of the entire manuscript is small, a high recognition rate can be obtained for the entire manuscript.

It is also possible to select not a single character area having a large area, a large number of lines or a large number of characters but a predetermined number of two or more, and determine the threshold value based on the density histograms thereof. Embodiment of the present invention.

Next, the operation of the mode is the same as that of the mode except that the character area is replaced with the table area.
The operation of the mode is the same as that of the mode except that one table area having the largest area is used instead of the character area for determining the threshold value. Modes and modes are suitable for handling documents with a high weight in the table area, such as a document with only a table and its caption.

<Third Embodiment> FIG. 7 is a schematic block diagram of an image processing apparatus according to the present embodiment. In FIG. 7, the same reference numerals as those in FIG. 3 indicate the same parts or corresponding parts, and the description thereof will be omitted. The difference from the second embodiment shown in FIG. 3 is that other areas, which are not the target of character recognition, are binarized by simple binarization processing in the same multi-value / binary conversion unit 112 as the character area and the table area. It is only the point that I tried to do. This multivalue-
The binarization threshold value of the binary conversion unit 112 is set by the threshold value control unit 120, but it has the same modes as the modes 1 to 3 in the second embodiment, and the user can specify any of the modes. The multi-value / binary conversion unit 111 is always valid during the binarization process.

In this embodiment, image data of a region such as a photograph or a halftone dot image is converted into a binary representation by a simple binarization method using a single threshold, so that it is compared with the case of dithering or the like. Inevitable deterioration of image quality such as photographs and halftone images. However, when this apparatus is used as a pre-processing apparatus for a character recognition apparatus, since the area excluding the characters and the table is originally outside the object of character recognition, the quality deterioration does not have a particularly bad effect on the character recognition performance.

<Fourth Embodiment> FIG. 8 is a schematic block diagram of an image processing apparatus according to the present embodiment. This apparatus adds a threshold control unit 130 having the same configuration as the threshold control unit 120 of the second embodiment shown in FIG. 3 to the configuration of the first embodiment shown in FIG. The configuration is such that the binarization thresholds of the multi-value / binary conversion units (simple binarization processing units) 108 and 109 are controlled.

This apparatus has the following modes (10) to (16) for controlling the binarization threshold of the character / table area, and the user can specify any mode.

Mode (10): This mode corresponds to the mode of the second embodiment. For each character area, the threshold for each character area is determined based on the density histogram of the area, and for each table area, the threshold for each table area is determined based on the density histogram of the area. The multi-value to binary conversion unit 108 is set, and the threshold for the table area is set to the multi-value to binary conversion unit 109. This mode and the mode described later (1
In 1), (12), (13), and (16), it is possible to specify whether to exclude solid white or black areas when obtaining the density histogram of the character area.

Mode (11): This mode corresponds to the mode of the second embodiment. A threshold is determined based on the density histogram of all character areas, and this threshold is set to the multi-value / binary conversion unit 10.
It is set for 8,109.

Mode (12): Corresponds to a mode obtained by combining the modes of the second embodiment with the modes. That is, the threshold value for the character area is determined based on the density histogram of all the character areas, this is set for the multi-value / binary conversion unit 108, and the threshold value for the table area is set based on the density histogram of all the table areas. Is determined and set in the multi-value / binary conversion unit 109.

Mode (13): This mode corresponds to the mode of the second embodiment. One character area having the largest area, number of lines or number of characters is selected, a threshold value is determined based on the density histogram, and the threshold value is set to the multi-value / binary conversion unit 108.
Set to 109.

Mode (14): This mode corresponds to the mode of the second embodiment. A threshold value is determined based on the density histograms of all the surface areas, and this threshold value is set to the multi-value / binary conversion unit 10.
Set for 8,109.

Mode (15): This mode corresponds to the mode of the second embodiment. Select one table area with the largest area, determine the threshold value based on the density histogram,
This threshold is set in the multi-value / binary conversion units 108 and 109.

Mode (16): Corresponds to the mode by combining the modes of the second embodiment with the modes. Select one character area with the maximum area, number of lines or number of characters, determine the threshold value based on the density histogram, and use this as a multi-value
It is set in the binary conversion unit 108. Further, one table area having the largest area is selected, a threshold value is set based on the density histogram, and this is set in the multi-value / binary conversion unit 109.

The operation in each of the above modes can be easily understood from each of the above-mentioned embodiments, and therefore the description will not be repeated here. According to the present embodiment, there are modes (10), (12), and (16) for controlling the binarization thresholds of the character area and the table area separately, so that compared to the second embodiment or the third embodiment, It is possible to binarize the image data of the character area and the table area with higher accuracy.

[0065]

As described in detail above, claim 1 is as follows.
According to the invention described in each of the fifth to fifth aspects, the binarization of the image data is performed by the same processing method or the simple binarization processing with a constant binarization threshold for each area in which the characters are written on the document. Can be stabilized. Claim 3 or 4
According to the described invention, the binarization threshold value is optimized based on the density histogram of the area where the character is written, so that the area where the character is written can be stably and highly accurately binarized. According to the invention described in claim 5, even if there is a white or black solid area inside the area in which characters are written, the influence thereof is eliminated and the binarization threshold value is optimized to perform highly accurate binarization. Can be done. Further, according to the invention as set forth in claim 4, since a part of the region is selected for determining the binarization threshold and the density histogram thereof is used, the processing amount for determining the binarization threshold is reduced. If the area where the number of characters is estimated to be high can be selected and the area where the number of characters is estimated to be large is selected, the binarization threshold value can be sufficiently optimized to perform highly accurate and stable binarization. If the image processing apparatus according to the present invention is used as a preprocessing apparatus for a character recognition apparatus, highly accurate and stable binarized data can be input to the character recognition apparatus in a region where a character is written on a document. Therefore, a significant improvement in recognition rate can be expected.

[Brief description of drawings]

FIG. 1 is a schematic block diagram showing a device configuration of a first embodiment.

FIG. 2 is an explanatory diagram of area identification.

FIG. 3 is a schematic block diagram showing a device configuration of a second embodiment.

FIG. 4 is a block diagram showing a configuration of a threshold control unit.

FIG. 5 is a diagram showing an example of a character area including a white solid portion.

FIG. 6A is a diagram showing an example of a density histogram of a character area. (B) It is a figure which shows the example of the density histogram of the area | region of a photograph or a halftone image.

FIG. 7 is a schematic block diagram showing a device configuration according to a third embodiment.

FIG. 8 is a schematic block diagram showing a device configuration of a fourth embodiment.

[Explanation of symbols]

Reference Signs List 100 multi-valued representation image data 102 image memory 104 area identification section 106 control section 108 multi-valued / binary conversion section for character area (simple binarization processing section) 109 multi-valued / binary conversion section for table area (simple Binarization processing unit) 110 Multi-value to 2-value conversion unit for other areas (dither processing unit) 111 Multi-value to 2-value conversion unit for character / table areas (Simple binary conversion processing unit) 112 For all areas Multi-value to binary conversion unit (simple binary conversion processing unit) 120 Threshold control unit 122 Solid part removal unit 123 Histogram creation unit 124 Threshold calculation unit 125 Threshold setting unit 130 Threshold control unit 200 Input document 202 to 205 Character area 206 Table Area 208,209 Separator 210 Figure 230 Character area 231 to 235 Character line 236,237 White solid part

Claims (5)

[Claims] 1. An area discriminating means for globally discriminating an area in which characters are written and other areas on the original from multivalued image data of the original, and a character discriminated by the area discriminating means. Multi-valued representation image data of the written area,
An image processing apparatus comprising: a multi-value / two-value conversion means for converting into binary expression image data by a constant processing method. 2. The image processing apparatus according to claim 1,
The multi-value to binary conversion means is a simple binary with a constant binarization threshold.
An image processing apparatus, which converts multi-valued expression image data into binary expression image data by a binarization process. 3. An area identifying means for globally identifying an area in which characters are written on the original and other areas from multivalued image data of the original, and a character identified by the area identifying means. Multi-value / binary conversion means for converting multi-valued expression image data in the written area into binary expression image data by simple binarization processing, and an area in which the character identified by the area identification means is written And a means for determining a binarization threshold value for the multivalue-to-binary conversion means based on the density histogram, and the binarization threshold value is at least the same. An image processing device characterized in that it is fixed in the area. 4. An area discriminating means for globally discriminating an area in which characters are written and other areas on the original from multivalued image data of the original, and a character discriminated by the area discriminating means. Multi-valued representation image data of the written area,
The multi-value / binary conversion means for converting into binary representation image data by the simple binarization processing and a part of the area in which the character identified by the area identification means is written are selected, A density histogram is obtained from the multivalued image data of the selected region, and a binarization threshold value for the multivalue-binary conversion means is determined based on the density histogram. Is fixed in at least the same area. 5. The image processing apparatus according to claim 3 or 4, wherein the means for determining the threshold value obtains a density histogram by excluding white or black solid portions inside the area.