JP2002142101A - Image processing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image processing device which is capable of accurately recognizing the direction of a manuscript through simple constitution even if the manuscript contains a color image including a halftone background. SOLUTION: The signals of each color of R, G, and B of image data are binarized through a binarization processor 207. When the character image of the binarized image is represented by white pixels, an image inversion process is carried out through an image inversion processor 210. The binarized image of each channel of R, G, and B is subjected to an AND operation to eliminate only the background, and a histogram is formed of only the character images. The angle of rotation of the image is recognized by a rotational angle recognition part 212 from the histogram, the image is rotated by an image rotating part 213, and manuscript image data are outputted as they are set coincident with each other in direction.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to an image processing apparatus applied to an image forming machine such as a copying machine. In particular, the present invention
The present invention relates to a measure for improving the reliability of the direction recognition (recognition of the paper direction) of a document having a background image or the like.

[0002]

2. Description of the Related Art Conventionally, a copier equipped with an automatic document feeder has been known. When a plurality of originals are continuously copied by this copying machine, for example, the plurality of originals placed on a paper feed tray are taken out one by one by an automatic original feeder, and an image is read. The image forming operation on the recording paper is sequentially performed based on this.

In this case, it is necessary for the user to check whether the originals placed on the paper feed tray are uniformly oriented in the same direction. If this checking operation is neglected, and if a plurality of originals have different directions (upside down), a copy of the original will be copied in a different direction. After the operation is completed, the user has to find a copy in a different direction and correct the direction. In particular, when a plurality of copies are made at the same time, a complicated operation of correcting a copy having a different direction is required for each copy set. Furthermore, when the copying machine is provided with an automatic stapling unit and the automatic stapling of the copy is performed, the stapling of the copy is performed in a different direction (upside down with the other copy). In the worst case, copying may need to be performed again.

In order to solve such a problem, even if the user does not check the orientation of the original, the orientation of the original is recognized based on the read original data, and the image output is set to an appropriate direction. In other words, Japanese Patent Application Laid-Open No. 9-9040 discloses a method of performing a copy operation by inverting image data of a document when a part of the document is reversed.

[0005]

As a conventional method for recognizing the orientation of a document, as disclosed in the above publication, a histogram indicating the density distribution of a document image is generated and used. Generally, the direction of the document is recognized. However, this technique requires accurate generation of the histogram. For example, when a document is scattered with a lot of noise such as minute dirt or dust, or when reading a document image including a halftone background, the line spacing or the beginning of a sentence may be affected by the noise or background. It may not be possible to determine the exact position such as For this reason, the generated histogram also lacks accuracy, and the reliability of document direction recognition is reduced.

To solve this problem, Japanese Patent Laid-Open Publication No.
There is an image recognition device disclosed in JP-A-00-22899. In this publication, the size of a character image is detected, and the filter size of an isolated point removal filter is determined in accordance with the detected character size, thereby efficiently removing the noise and performing accurate document direction recognition. The technical idea is disclosed.

According to the technology disclosed in this publication,
It is possible to efficiently remove isolated point noise and the like with a simple configuration. However, in the case of a color image including a halftone background as shown in FIG. 12A, when there is no large difference between the luminance of the character color and the luminance of the background, or when the luminance of the background is not uniform, It is highly likely that character detection from the histogram becomes difficult. FIG. 12B shows two histograms obtained from the image shown in FIG.
It is a valued image. Thus, in the conventional technology,
Character recognition cannot be performed satisfactorily on a color image including the background as described above, and the reliability of document direction recognition cannot be sufficiently obtained.

As a means for removing the above noise and background, there is a method proposed in Japanese Patent Application Laid-Open No. 7-99581. However, this method requires enormous processing such as recognition processing using a neural network and vector calculation. However, it takes a long time for processing, and thus lacks practicality.

SUMMARY OF THE INVENTION The present invention has been made in view of the foregoing, and an object of the present invention is to provide a simple structure for accurately recognizing the original direction even for a color image including a halftone background. It is to provide a possible image processing device.

[0010]

Means for Solving the Problems Summary of the Invention In order to achieve the above object, the present invention performs a binarization process on each of RGB signals of image data,
By synthesizing the obtained image data, the background image is removed, and only the character image is obtained so that the original direction can be accurately recognized.

-Solution Means- Specifically, an image processing apparatus for recognizing the orientation of a document including a character image is assumed. The image processing apparatus is provided with an area separating unit, a binarizing unit, a combining unit, a rotation angle recognizing unit, and a rotating unit. The area separating means separates the image data read from the document into a character area and a non-character area. The binarization processing means performs a binarization process on each of the RGB signals of the image data. The combination processing means
The image data of each of the RGB colors subjected to the binarization process is synthesized. The rotation angle recognizing means recognizes the rotation angle of the image based on the image data subjected to the synthesis processing. The rotation processing means rotates image data according to the rotation angle of the image.

[0012] According to this specific matter, the image data obtained by performing the binarization processing on each of the RGB signals is combined, the background image is removed, and only the character image is obtained. Then, the rotation angle of the image is recognized based on the character image, and the image is rotated if necessary. For example, if the document is upside down, the image is rotated by 180 ° to match the direction of the image with the direction of another document image. As a result, even if the originals do not all face the same direction, the directions of the image data of all the originals match.

The image processing apparatus further includes resolution conversion means for converting image data of at least a character area of the character area and the non-character area separated by the area separation means to a low resolution. When the binarization process is performed on the converted image data,
By reducing the amount of data accompanying the reduction in resolution, processing can be speeded up and resources can be saved.

The following is a specific configuration of the synthesis processing means. That is, the synthesizing processing means synthesizes each image data which has been binarized by the binarizing processing means with respect to the RGB signals for each color by a logical product operation. With this specific matter, it is possible to represent all of the character image with black pixels, and it is possible to realize a highly reliable combination process that can reliably extract only the character image.

A histogram generation means for generating a histogram indicating the density distribution of the original image from the image data, and an image shape of the image data binarized by the binarization processing means are determined based on the histogram. An image shape determining unit that determines whether or not the character image of the binarized image data is represented by white pixels based on the average density and the image shape of the document image, and the character image is represented by white pixels. If it is determined that the character image is represented by white pixels, the image inversion processing means for performing inversion processing on the binarized image is provided. Images can be represented by black pixels. For this reason, extraction of only a character image can be performed more reliably. In other words, looking at each of the RGB channel data in the color image, it is assumed that the luminance of the character image is uniform in any channel, while the case of the halftone background image is uniform and non-uniform. In any case, the luminance of the background image of each of the R, G, and B channels is rarely the same as that of the character image, and even if the luminance is the same, the background image is often scattered as isolated point noise. The present invention binarizes each of the RGB channels individually, determines whether the character image of the binarized image is represented by white pixels, and determines whether the character image is represented by white pixels. Image inversion processing is performed on the value-processed image. RG on it
By performing an AND operation on the binarized image of each channel B, it is possible to remove only the background and create an accurate histogram of only the character image.

The following is a specific example of a method for determining an image shape using a histogram. That is, based on the histogram generated by the histogram generation means, the edge shape of the histogram is detected,
An edge counting means for counting the number of edges and an edge number storage means for storing the number of edges are provided, and the image shape determination means determines an image shape based on the stored edge number. I have. According to the specific items, the image shape can be accurately determined, and the rotation angle of the image based on the character image can be more accurately recognized.

The specific configuration of the rotation angle recognition means is as follows. That is, the rotation angle recognizing means recognizes the direction of the character image based on the histogram indicating the density distribution of the document generated from the image data from which the halftone background and the non-character area have been removed. With this specific matter, it is possible to recognize the direction of the character image with high efficiency and accuracy.

[0018]

Embodiments of the present invention will be described below with reference to the drawings. In this embodiment, a case will be described in which the present invention is applied to a digital color copying machine having an automatic document feeder.

FIG. 1 is a block diagram showing a configuration for performing image processing of the color image copying machine according to the present embodiment. As shown in FIG. 1, the present color image copying machine includes a color image input device 1, an image processing device 2, and a color image output device 3. Further, the image processing apparatus 2 includes an A / D conversion unit 21, a shading correction unit 22, an input gradation correction unit 2,
3, an image processing unit 24, a color correction unit 25, a black generation and under color removal unit 26, a spatial filter processing unit 27, an output gradation correction unit 28, and a gradation reproduction processing unit 29.

The color image input device 1 is, for example, a CCD
(Charge Coupled Device). Then, the image of the reflected light of the light irradiated toward the original is converted into RGB (R: red, G: green,
B: blue) is read by the CCD as an analog signal,
This analog signal is input to the image processing device 2.

The analog signals read by the color image input device 1 are transmitted to respective units 21 to 29 in the image processing device 2.
, And predetermined image processing described later is performed.
It is output to the color image output device 3 as digital color signals of M, Y, K (C: cyan, M: magenta, Y: yellow, K: black). Then, an image is formed on a sheet based on the color signal. Hereinafter, each unit 21 to 29 in the image processing apparatus 2 will be described.

The A / D converter 21 converts the RGB analog signals into digital signals.

The shading correction unit 22 performs a process for removing various types of distortion generated in the illumination system, the imaging system, and the imaging system of the color image input apparatus 1 from the RGB digital signal received from the A / D conversion unit 21. It is something to give.

The input tone correction section 23 outputs an RGB signal (R) from which various distortions have been removed by the shading correction section 22.
At the same time as adjusting the color balance of the (GB reflectance signal), a process of converting the signal into a signal such as a density signal that is easy to handle by the image processing system employed in the image processing apparatus 2 is performed.

The image processing section 24 includes an area separation processing section 202 as area separation means and a rotation angle recognition section 212 (see FIG. 2 showing the internal configuration of the image processing section 24). The area separation processing unit 202 separates the RGB signals by determining whether each pixel in the input image belongs to a character area, a halftone area, or a photograph area. Rotation angle recognition unit 212
Is to determine the top and bottom of the input image data (vertical direction of the image) in order to perform the rotation processing of the image data so that the image output is in an appropriate direction. The image processing unit 2
4 outputs an area identification signal indicating to which area the pixel belongs based on the area separation result,
After being output to the spatial filter processing unit 27 and the tone reproduction processing unit 29, respectively, the input signal received from the input tone correction unit 23 is subjected to a rotation process as necessary, and then sent to the subsequent color correction unit 25. Output.

The black generation and under color removal unit 26 performs the color-corrected CM
A black generation process for generating a black (K) signal from the three Y color signals, and a process for generating a new CMY signal by subtracting the K signal obtained by the black generation from the original CMY signal. The black generation and under color removal unit 26 converts the CMY three-color signals into CMYK four-color signals.

As an example of the black generation processing, there is a method of generating black using skeleton black (a general method of black generation processing). In this method, the input / output characteristics of the skeleton curve are set to y = f (x), and the input data is C, M,
Y, the output data are C ', M', Y ', K', and the UCR (Under Coor Removal) rate is α (0
When <α <1), the black generation / undercolor removal processing is represented by the following equation.

K ′ = f {min (C, M, Y)} C ′ = C−αK ′ M ′ = M−αK ′ Y ′ = Y−αK ′ The spatial filter processing unit 27 removes black generation and undercolor. The image data of the CMYK signal received from the unit 26 is subjected to a spatial filter process using a digital filter based on the region identification signal to correct the spatial frequency characteristics so as to prevent "blur" and deterioration of the granularity of the output image. To be processed. The output tone correction unit 28 and the tone reproduction processing unit 29 also
Similarly to the spatial filter processing unit 27, predetermined processing is performed on the image data of the CMYK signal based on the area identification signal.

For example, in order to enhance the reproducibility of a black character or a color character, a region separated as a character in the region separation processing unit 202 is emphasized at a high frequency by a sharp enhancement process in the spatial filtering process by the spatial filtering unit 27. The amount is set large. At the same time, the tone reproduction processing section 29 selects binarization or multi-value processing on a high-resolution screen suitable for reproduction of a high frequency band.

Further, for the area separated as a halftone dot by the area separation processing section 202, the spatial filter processing section 27 performs a low-pass filter processing for removing an input halftone dot component. Then, the output gradation correction unit 2
In step 8, an output tone correction process for converting a signal such as a density signal into a dot area ratio, which is a characteristic value of the color image forming unit, is performed. A tone reproduction process (intermediate tone generation) is performed to separate and reproduce each tone.

Further, with respect to the area separated as a photograph in the area separation processing section 202, a binarization or multi-value processing is performed on a screen which emphasizes tone reproducibility.

The image data subjected to each of the above-described processes is
The data is temporarily stored in a storage unit (not shown), read out at a predetermined timing, and input to the color image output device 3.

This copying machine forms an image on a recording medium (for example, paper) based on image data.
It is configured as a color image forming machine using an electrophotographic system or an inkjet system. Further, the image forming machine to which the present invention is applied is not limited to a copying machine,
It may be a printer or a facsimile machine.

FIG. 2 shows the detailed configuration of the image processing section 24.
Hereinafter, the processing procedure of the image data in the image processing unit 24 will be described with reference to the block diagram of FIG. 2 and the flowcharts of FIGS.

The input tone correction section 23 sends the image processing section 2
RGB signals input to the high-resolution image memory 201
Is stored once. FIG. 11A shows the above-mentioned FIG.
3 shows respective images based on the RGB signals obtained from the original image. Thereafter, the high-resolution image memory 20
1 is read out, the area separation processing unit 202 separates the area into a character area and a non-character area (halftone area and photograph area), and extracts only the character area (step ST1 in FIG. 3). As described above, by processing only the character area, the amount of data is reduced, so that the processing can be speeded up and resources can be saved.

Note that the region separation here is performed using a known technique. For example, "Presentation of the Institute of Image Electronics Engineers of Japan 90-0
6-04 ".

Here, the following determination is performed in a block of M × N pixels (M and N are natural numbers) centered on the target pixel (pixel to be processed), and this is used as a region identification signal of the target pixel. .

An average value (Dave) of the signal level is obtained for the central nine pixels in the block, and each pixel in the block is binarized using the average value. In addition, the maximum pixel signal level (Dmax) and the minimum pixel signal level (Dmin) are determined at the same time. In the halftone dot area, the fluctuation of the image signal in the small area is large, and the density is higher than the background. Utilizing this, the halftone dot area is identified. For the binarized data, the number of change points from 0 to 1 and the number of change points from 1 to 0 are obtained in the main scanning and sub-scanning directions, respectively, and K _H and K _V are obtained, respectively.
And then, by comparing the threshold T _H, and T _V, when each change points are both greater than the threshold, it halftone dot region. In order to prevent erroneous determination with the background, the above Dmax, Dmin,
Dave is compared with threshold values B ₁ and B ₂ .

Dmax−Dave> B ₁ , Dave−Dmin> B ₂ , K _H > T _H , and K _V > T _V ... Judgment as a halftone area Other than the above conditions... Judgment as a non-halftone area Since the difference between the maximum signal level and the minimum signal level is large and the density is considered to be high, the character area is identified as follows. That compares maximum was obtained previously in the non-halftone dot region, a minimum signal level and their difference (Dsub) threshold P _A, P _B, and P _C, one of the maximum, minimum signal level and the difference If one exceeds the threshold, it is a character area, and if all of them are below the threshold, it is a photographic area.

Dmax> P _A , or Dmin> P _B , or Dsub> P _C ... Judgment as a character area Other than the above conditions. Judgment as a photographic area Characters subjected to area separation processing by the area separation processing unit 202 as described above The data of the area portion is resolution-converted to a low resolution by a resolution conversion unit 203 as resolution conversion means. For example, data read at 400 DPI is stored in the low-resolution image memory 204 as low-resolution image data such as 25 DPI. High-resolution data is not required because the height of the character image can be detected to create a histogram for document orientation recognition, so high-speed processing and resources can be achieved by converting to low-resolution data and reducing the amount of data. It is possible to save money.

Binarization processing section 20 as binarization processing means
In step 7, data in the low-resolution image memory 204 is read, and binarization processing is performed on each of the RGB signals (step ST2 in FIG. 3). The binarization procedure at this time is performed by comparing the gradation data for each pixel with a threshold for binarization (threshold stored in the threshold storage unit 217 in FIG. 2).

Here, the setting of the threshold value at the time of performing the binarization processing is performed using a known technique. For example, a density average value of an image is obtained, and this is set as a threshold. Rather than using the average of the densities, some calculation may be performed on the average value to set the threshold. FIG. 11B shows images obtained by binarizing the RGB channel images.

On the basis of the binarized image, a histogram showing the density distribution of the original image is created by the histogram creating section 205 as a histogram creating means (step ST3 in FIG. 3). The histogram is generated by counting the number of black pixels in the main scanning direction or the sub-scanning direction. The binarized images are classified into four patterns according to the setting of the threshold value and the combination of the character color and the background color, and are created. The histogram is also classified into four patterns. One is where the character image and the background image are separated and the character image is represented by black pixels (see FIG. 9A). One is where the character image and the background image are separated and the character image is represented by white pixels. In the case (see FIG. 9B), one is when the character image and the background image are not clearly separated and the character image is represented by black pixels (see FIG. 9D). Are not clearly separated and the character image is represented by white pixels (see FIG. 9C).

When the luminance of the character color is low and the luminance of the intermediate background color is generally high, the character image and the background image are separated, and the character image is converted into black pixels. Conversely, if the luminance of the character color is high and the luminance of the intermediate background color is low overall, the background image and the character image are separated, and the character image is converted to white pixels.

In these two patterns, when the character image is represented by black pixels, the average density of the image is low, and when the character image is represented by white pixels, the average density of the image is high.

When the difference between the luminance of the intermediate color of the background and the luminance of the character color is relatively small, the background image and the character image cannot be accurately separated. When the background image portion largely remains, the number of histogram edges (change points) counted in the histogram edge (change point) count process (FIG. 6) decreases. In this case, the character image can be roughly classified into a case where the character image is represented by black pixels and a case where the character image is represented by white pixels.

In the two patterns, on the contrary, when the character image is represented by black pixels, the average density of the image is high, and when the character image is represented by white pixels, The average density will be lower.

Next, the image shape discriminating process for discriminating the shape of the histogram is performed (step ST4 in FIG. 3).

The image shape discriminating process performed by the image shape discriminating unit 209 as the image shape discriminating means will be described below with reference to the flowchart of FIG.

The main scanning direction edge counting process (step ST11 of FIG. 4) and the sub-scanning direction edge counting process (step ST11 of FIG. 4) are performed on the generated histogram.
In step 12), the number of edges (change points) of the histogram is counted. FIG. 6 shows the histogram edge counting procedure.
Is shown in the flowchart of FIG.

At the time of the counting process, the comparison target memory 21
5 is initialized (for example, 0 if the initial point is white, 1 if it is black) (step ST31 in FIG. 6). Next, n, which is a number for rearranging the appearance order of the change points of the histogram, is set to “1” (step ST32 in FIG. 6).

The data stored in the comparison target memory 215 is compared with the change point data of the n-th histogram (step ST33 in FIG. 6) (step ST34), and the value of the change point data of the histogram is larger than that of the comparison target data. If it is larger than the value of the threshold storage unit 206 (step ST3)
4, the histogram edge counter 21 serving as an edge counting means assuming that the edge is a rising edge.
4 is incremented (step ST
35). The counted number of edges is stored in the edge number storage unit 216 as edge number storage means. The change point data of the histogram at this time is stored in the comparison target memory 215 (step ST36). If it is not regarded as an edge (NO in step ST34), the histogram change point data at the next change point is read (steps ST42 and ST33), and the same comparison as above is performed. In the case of these two patterns, since the rising edge and the falling edge appear alternately as shown in FIG. 9, only the case where they appear alternately is counted. Then n
Is incremented (steps ST43 and ST3).
7) The next change point data of the histogram is read (step ST38), and the read change point data of the histogram is compared with the comparison target data (step ST39).
If the data to be compared is larger than the histogram data by more than the threshold (YES in step ST39), the value of the counter is regarded as a falling edge and the value of the counter is incremented (step ST40). The value of the change point data of the histogram at this time is stored in the comparison target memory 215 (step ST41). If it is not regarded as an edge (NO in step ST39), the value of n is incremented (step ST37), the change point data of the histogram is read (step ST38), and the same comparison as above is performed. The number of rising edges, the number of falling edges,
When all data processing to be performed is completed in any of the counts (when the count processing for all the change points is completed), the edge count processing is completed (YES determination in steps ST43 and ST44).

In the present invention, the predetermined value of the threshold value storage unit 206 is set to “16”. This predetermined value is not limited to this value.

Returning to the flowchart of FIG. 4, in the image shape discriminating process, the number of edges in the main scanning direction and the number of edges in the sub-scanning direction are compared, and the larger one is subjected to the next process (step in FIG. 4). ST13). When the number of edges in the main scanning direction or the sub-scanning direction is larger than a predetermined value (YES in step ST14 or step ST15), that is, when the character image is one of the four patterns in which the character image can be accurately separated, the edge The determination flag is set to “0” (step ST16). On the other hand, when the number of edges in the main scanning direction or the sub-scanning direction is smaller than a predetermined value (step ST14 or step ST15).
If the character image and the background image are not separated, the edge determination flag is set to "1" (steps ST17 and ST18).

Returning to the flowchart of FIG. 3, when it is determined that the character image is represented by white pixels, the image inversion processing section 210 as image inversion processing means performs image inversion processing and converts the character image into black pixels. (Step ST5 in FIG. 3). The image inversion procedure at that time is shown in the flowchart of FIG. First, if the edge discrimination flag is "1" (YES in step ST21 of FIG. 5), the inversion flag is turned on (step ST22). Conversely, if the flag is "0" (NO in step ST21), The inversion flag is turned off (step ST23). Then 2
The average density of the binarized image is obtained and compared with a predetermined value. If the average density of the binarized image is larger than a predetermined value (predetermined density) (YES in step ST24), the inversion flag is inverted (step ST25). ).

The inversion flag is turned ON by the above processing when the edge determination flag is "1" and the average density of the image is lower than a predetermined value, and when the edge determination flag is "0" and the average density of the image is predetermined. It is higher than the value. The former is a case where the character image is separated and the character image is represented by white pixels (see FIG. 9B), and the latter is a case where the character image is not separated and the character image is represented by white pixels (see FIG. c)). If the inversion flag is ON (YES in step ST26), the image is inverted (step ST27).
The above is the image inversion processing procedure.

Next, the processing from step ST2 to step ST5 in FIG. 3 is sequentially performed on the image data of each of the RGB channels, and the data is stored in the low-resolution image memory 204. The stored data is subjected to logical product synthesis in the image synthesis unit 211 as a synthesis processing unit (step ST6 in FIG. 3). That is, in the image combining unit 211,
Each image data obtained by binarizing the RGB signals is combined by a logical product operation. Since the character images are all represented by black pixels in the processing so far, the character image portion is represented by black pixels even in the output of the logical product of three inputs and one output. On the other hand, since the background image is represented by white pixels or black pixels in each of the RGB channels, the logical product output is represented by white pixels by performing an appropriate inversion process. FIG. 11C shows an image obtained by synthesizing the above images.

Next, the rotation angle recognition section 212 as rotation angle recognition means recognizes the rotation angle of the read image data (step ST7 in FIG. 3). The image rotation angle recognition processing is shown in the flowchart of FIG. First, a histogram of the combined output image data is created, and its edges (change points) are counted (steps ST51 and S51 in FIG. 7).
T52). The histogram edge (change point) counting means is shown in the flowchart of FIG. Here, unlike the edge counting process described with reference to FIG. 6, the rising edge and the falling edge are separately counted. The edge count processing procedure in FIG. 8 is substantially the same as that in FIG. Steps ST61 to ST in FIG.
T70 corresponds to steps ST31 to ST3 in FIG.
4, ST39, ST35, ST40, ST36, ST4
3, ST42.

Next, the number of edges in the main scanning direction and the number of edges in the sub-scanning direction are compared (step ST53 in FIG. 7). The image stored in the high-resolution image memory 201 is shown in FIG.
In the case where the orientation is as shown in (a), in order to perform copy output in a state where the top and bottom are in a normal state, the output may be performed without performing the rotation process. In this case, the number of edges in the main scanning direction is larger than the number of edges in the sub-scanning direction (Y in step ST53).
ES determination), the number of falling edges increases in the histogram in the sub-scanning direction (YES determination in step ST59). If the image stored in the high-resolution image memory 201 has the orientation shown in FIG. 10B, a 180 ° rotation process is performed on the image in order to perform copy output in a normal state. Good. In this case, the number of edges in the main scanning direction is larger than the number of edges in the sub-scanning direction (step ST5).
3, the rising edges increase in the histogram in the sub-scanning direction (YES in step ST58). When there is not much difference between the number of rising edges and the number of falling edges (NO in step ST59), the rotation direction is set to 0 ° because the document direction cannot be determined. Similarly, in the case of the orientation shown in FIG. 10C, a 90 ° rotation process may be performed. In this case, the number of edges in the sub-scanning direction is larger than the number of edges in the main scanning direction (step ST5).
No. 3), the number of falling edges increases in the histogram in the main scanning direction (YES in step ST57). Further, in the case of the direction shown in FIG. 10D, the number of edges in the sub-scanning direction is larger than the number of edges in the main scanning direction (NO in step ST53), and the number of rising edges in the histogram in the main scanning direction is larger. (YES determination in step ST56). Similarly, if there is not much difference between the number of rising edges and the number of falling edges, the direction of the original cannot be determined, so the rotation angle is set to 0 ° (NO in step ST57).

In the present invention, step ST5 in FIG.
6, the predetermined value A used for determination in ST58 is "2", and the predetermined value B used for determination in steps ST57 and ST59.
Was set to "-2". These predetermined values are not limited to these values.

The rotation processing section 213 as the rotation processing means performs the rotation processing of the rotation angle set by the rotation angle recognition section 212 on the image stored in the high resolution image memory 201 as described above (FIG. 3). Step ST8).

As described above, in this embodiment, the rotation angle of the image is recognized, and the rotation processing of the image data is performed according to the rotation angle. Therefore, even if a part of a plurality of documents placed on the automatic document feeder is reversed, the directions of the image data of all the documents can be matched, and all the copies can be copied. It is possible to perform the copying operation in a state where the directions are adjusted. This allows the user to check whether the originals placed on the automatic document feeder are uniformly oriented in the same direction, or to search for a copy in a different direction after the copy operation and to change the orientation. Correction work is not required, and the burden on the user can be reduced.

[0063]

As described above, in the present invention, each of the RGB signals of the image data is subjected to the binarization process, and the image data obtained thereby is synthesized, whereby the background image is obtained. Thus, the document direction can be accurately recognized by obtaining only the character image. For this reason, the background image and the non-character area can be removed with a simple configuration, and accurate document direction recognition can be performed. As a result, the work load on the user can be reduced, and the usability of the apparatus can be improved with a relatively simple configuration.

When the image data in the character area is converted to a low resolution and the binarization process is performed, the amount of data accompanying the low resolution is reduced, thereby increasing the processing speed and saving resources. It becomes possible to plan.

When the binarized image data are combined by a logical product operation, all of the character images can be represented by black pixels, and the intermediate configuration can be easily performed with a simple configuration. A tonal background can be removed, and a highly reliable synthesis process that can reliably extract only a character image can be realized.

Further, a histogram indicating the density distribution of the original image is generated from the image data, the image shape of the image data is determined based on the histogram, and if the character image is represented by white pixels, the binary When the inversion processing is performed on the coded image, even if the character image is represented by white pixels, the character image can be represented by black pixels. For this reason, the extraction of only the character image can be performed more reliably, and the reliability of document direction recognition can be improved.

In addition, when the edge shape of the histogram is detected based on the generated histogram and the image shape is determined based on the number of edges, the image rotation based on the character image is performed. Corners can be more accurately recognized.

Furthermore, when the direction of the character image is recognized based on the histogram indicating the density distribution of the document, the direction of the character image can be recognized with high efficiency and accuracy.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of a color image copying machine according to an embodiment.

FIG. 2 is a block diagram illustrating a detailed configuration of an image processing unit.

FIG. 3 is a flowchart illustrating a document direction recognition processing operation.

FIG. 4 is a flowchart illustrating an image shape determination processing operation.

FIG. 5 is a flowchart illustrating an image inversion processing operation.

FIG. 6 is a flowchart illustrating a first histogram edge count processing operation;

FIG. 7 is a flowchart illustrating an image rotation angle recognition processing operation.

FIG. 8 is a flowchart illustrating a second histogram edge count processing operation.

FIG. 9 is a diagram for explaining a difference between a non-reversed document and a reversed document.

FIG. 10 is a diagram for explaining a difference between histograms according to a rotation angle of a document.

11A is a diagram showing RGB channel images of the image of FIG. 12A, FIG. 11B is a diagram showing a binary image of the channel image of FIG. 12A, and FIG. 11C is a diagram of FIG. FIG. 7 is a diagram showing an image obtained by combining the images of FIG.

12A is a diagram illustrating an example of an original color image of a document, and FIG. 12B is a diagram illustrating an image obtained by performing a binarization process on the image of FIG.

[Explanation of symbols]

2 Image Processing Apparatus 202 Region Separation Processing Unit (Region Separation Unit) 203 Resolution Conversion Unit (Resolution Conversion Unit) 205 Histogram Creation Unit (Histogram Generation Unit) 207 Binarization Processing Unit (Binarization Processing Unit) 209 Image Shape Discrimination Processing Unit (image shape discriminating unit) 210 image inversion processing unit (image inversion processing unit) 211 image synthesizing unit (synthesis processing unit) 212 rotation angle recognition unit (rotation angle recognition unit) 213 rotation processing unit (rotation processing unit) 214 edge counter (Edge counting means) 216 Edge number storage section (Edge number storage means)

Continued on the front page F-term (reference) AA25 AA27 BA03 BA06 CA10 CB01 5C077 MP05 MP08 PP13 PP20 PP22 PP23 PP25 PP27 PP32 PP46 PP47 PP57 PP59 PQ12 PQ17 PQ18 PQ19 PQ22 RR02

Claims (6)

[Claims] 1. An image processing apparatus for recognizing a direction of a document including a character image, comprising: a region separating unit that separates image data read from the document into a character region and a non-character region; 2 for each color signal
Binarization processing means for performing the binarization processing; synthesis processing means for synthesizing the image data of each of the RGB colors which have been subjected to the binarization processing; rotation for recognizing the rotation angle of the image based on the synthesized image data An image processing apparatus comprising: a corner recognition unit; and a rotation processing unit that performs rotation processing on image data according to a rotation angle of the image. 2. The image processing apparatus according to claim 1, further comprising: resolution conversion means for converting image data of at least a character area of the character area and the non-character area separated by the area separation means to a low resolution. An image processing device configured to perform a binarization process on the image data converted to the low resolution. 3. The image processing apparatus according to claim 1, wherein the synthesizing processing unit performs a logical AND operation on each image data that has been binarized by the binarization processing unit with respect to the RGB signals. An image processing apparatus characterized in that the image processing apparatus is configured to combine images. 4. The image processing apparatus according to claim 1, wherein a histogram indicating a density distribution of the document image is generated from the image data, and a binarization processing unit. Image shape discriminating means for discriminating the image shape of the image data binarized by the above based on the histogram; and character image of the binarized image data based on the average density and the image shape of the original image. Determining whether the character image is represented by a white pixel by determining whether or not the character image is represented by a white pixel; Image processing device. 5. An image processing apparatus according to claim 4, wherein an edge shape of the histogram is detected based on the histogram generated by the histogram generation means, and the number of edges is counted. An image processing apparatus comprising: edge number storage means for storing a number; and an image shape determination means configured to determine an image shape based on the stored edge number. 6. The image processing apparatus according to claim 1, wherein the rotation angle recognizing means includes a document generated from image data from which a halftone background and a non-character area have been removed. An image processing apparatus characterized in that the direction of a character image is recognized based on a histogram indicating the density distribution of the image.