JPH11272864A - Optical character reading method and device - Google Patents

Abstract

(57) [Summary] [Problem] To obtain a binary image for each purpose of reading, distinguished by color, from a form filled in with a plurality of colors without physically exchanging filters. SOLUTION: The image information of a form (1) is read and converted into an RGB signal (2), and the RGB signal is subjected to arithmetic processing to obtain a signal (HSV) relating to hue (H), saturation (S) and luminance (V). Signal), and the HSV signals are input to a plurality of HSV-density conversion means (4) having respective conversion function tables, and converted into respective density signals (4). The signal is binarized and converted into binary images of a plurality of systems (5), and for each system, a binary image having a different reading purpose such as a character recognition image by removing a dropout color, an extracted image of a specific color, etc. Getting images.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an optical character reading apparatus, and more particularly, to a form in which characters, backgrounds, character entry frames, notes, etc. are entered in two or more different colors.
The present invention relates to an optical character reader having a function of selectively reading necessary image information.

[0002]

2. Description of the Related Art In an optical character reading apparatus, a form in which a character entry frame is printed in a dropout color is used as a form to be read. The dropout color can be clearly distinguished by human eyes, but is a color having a hue and a density that cannot be distinguished from the background color by the sensor of the optical character reader.

[0003] If characters are written on a form in which a character entry frame is entered with a dropout color using a pencil, a ballpoint pen, or a printer, the character entry frame disappears when read by an optical character reader, and should be read. Image information of only characters can be extracted, and highly accurate character recognition becomes possible. Here, the character entry frame includes not only the literal frame itself, but also the one with the inside of the frame colored or the eight characters previously entered.

In the conventional optical character reading apparatus, in order to realize the dropout color, an optical filter having characteristics adapted to the dropout color and a writing implement (pencil or ballpoint pen) has been used. In this method, there is a problem that the filter has to be replaced depending on the color system of the dropout color (red system, blue system).
In addition, in order to correspond to a wide range of dropout colors, it is necessary to reduce the density, and there is a problem that a character entry frame and a note written in the dropout color become difficult to see.

[0005] This is because the optical filter has a characteristic that the transmittance of a specific wavelength is merely different. That is, the originally dark color does not lighten through the filter.

Some optical character reading devices are used for displaying on a screen and confirming them separately from image information dropped out for character recognition, and for saving a form image.
It has a function to simultaneously read image information captured without dropping out the character entry frame.

In the conventional optical character reading apparatus, in order to simultaneously capture a plurality of pieces of image information according to the purpose, a plurality of optical systems are provided, and a filter used in each optical system is switched. Such a device has a problem that the device becomes large because it has a plurality of optical systems.

[0008] In some cases, image information of a form is fetched as a color image, and a dropout process is performed based on the color information of the image. As one example, JP-A-2-67689 is cited. FIG. 3 is a principle block diagram of the technology described in the above-mentioned publication. In the image signal processing device described in the above publication, the form 1 is read as a three-color image signal, and is converted by the image information conversion means 2 into hue, saturation, and luminance information.

Here, hue (Hue) and saturation (Satu)
ratio) and luminance (Value) are expressed in the HSV color system. FIG. 1 illustrates the HSV color system. The saturation S indicates the vividness of the color. A value of 0 indicates an achromatic color, a larger value indicates a more vivid color (chromatic color), and a maximum value indicates a pure color. Hue H represents hue. In the present specification, the hue range is from 0 to 360 degrees, where 0 degrees represents red, 120 degrees is green, and 240 degrees is blue. The luminance V indicates brightness.

In the image signal processing apparatus shown in FIG. 3, the saturation S obtained by the conversion is compared with a predetermined threshold value Ts to determine whether the color is achromatic or chromatic (8). Is compared with a predetermined threshold Tv to determine whether the achromatic portion is a background or a character (9), and the hue information obtained by the conversion is converted to a predetermined threshold Tv.
The chromatic portion is output as multi-valued data in comparison with h (10). Here, it is assumed that the background is white (saturation = 0) and the characters are black (saturation = 0).

According to the above publication, it is possible to obtain an output obtained by dropping out a character entry box entered in chromatic colors without replacing a filter. Further, the character entry frame can be read out separately from the character by the output of the color determination means 10.

However, in the image signal processing apparatus disclosed in the above publication, since the saturation information, the luminance information, and the hue information are individually thresholded, there is a problem that the accuracy of dropout is reduced for the following reasons. is there.

In general, the conversion from the RGB color system to the HSV color system is expressed by the following equation.

[0014]

(Equation 1)

[0015]

(Equation 2)

[0016]

(Equation 3)

Note that r, g, and b are normalized in the range of 0 to 1, v and s are obtained in the range of 0 to 1, and h is obtained in the range of 0 to 360.

As shown by this equation, the luminance v is r, g,
b, the maximum component, the saturation s is determined by the ratio of the minimum component to the maximum component, and the hue h is determined by the ratio of the difference between the maximum component and the minimum component and the difference between the second component and the minimum component, centering on the maximum component. .

Thus, the saturation s and hue h are r,
Since it is determined by the ratio of each component of g and b, even if the luminance v is sufficiently small, that is, even if it is sufficiently close to black, r, g, b
If there is variation in each component, a value close to 1 can be obtained for the saturation s. For example, as a color almost black, r = 0.
1, g = 0, b = 0.1, v = 0.1, s = 1,
h = 300 is obtained. Here, the image signal processing device of the above publication determines whether a color is a chromatic color or an achromatic color only based on the value of the saturation s. Therefore, this color is determined as a chromatic color.

When a form is actually read as a color image, even if a character is written in achromatic black due to factors such as noise and color misregistration, a drop in the component may be taken into consideration. To determine whether the character is a character entry frame written in out-color or a character written in achromatic color, comprehensive judgment is required, not individual judgment of saturation, luminance, and hue.

In the image signal processing apparatus disclosed in the above publication,
Since the determination is made on the assumption that the character to be subjected to character recognition is written in achromatic color, the character written in chromatic color, for example, the second and subsequent characters of the copy paper (the first one is a black ballpoint pen) Is written in black letters, but the second and subsequent sheets are colored with blue ink and become blue letters, etc.). There is.

Here, the form image is left as a color image,
For example, if the image is captured as a full-color image and subjected to image processing by software, it is not difficult to solve the above problem. However, since a color image has a larger amount of data than a binary image, if data transfer / processing / storing is performed without changing the color image, the performance is reduced as compared with a conventional OCR system using a binary image. Would.

[0023]

SUMMARY OF THE INVENTION According to the present invention, the filter characteristics are changed without physically replacing the filter.
An object of the present invention is to provide an optical character reader having a function of obtaining a binarized image.

Further, the present invention makes it possible to handle forms that are difficult with the conventional method, such as dark dropout colors, extraction of specific colors, and reading of colored characters.
An object of the present invention is to provide an optical character reading device having a high degree of freedom in filter characteristics for obtaining a value image.

Further, the present invention provides an optical character reading apparatus which has the above-mentioned functions and does not cause a decrease in performance and a large increase in data amount.

[0026]

In order to solve the above problems, the present invention mainly employs the following configuration.

The image information of the form is read and converted into RGB signals, and the RGB signals are subjected to arithmetic processing to obtain signals (HS) relating to hue (H), saturation (S) and luminance (V).
V signal), the HSV signals are input to a plurality of systems of HSV-density conversion means having respective conversion function tables, and are converted into respective density signals. An optical character reading method for converting each image into a binary image having a different purpose of reading, such as a character recognition image by removing a dropout color, an extracted image of a specific color, or the like, for each system.

Further, image information conversion means for reading a form as an RGB image signal, and RGB for converting the RGB signal obtained by the image information conversion means into hue (H), saturation (S), and luminance (V). -HSV conversion means, and the RGB-HS
The apparatus further comprises at least one binarization unit for converting the HSV signal obtained by the V conversion unit into a binary image, wherein the binarization unit comprises:
An HSV-density conversion unit that converts the HSV signal obtained by the RGB-HSV conversion unit into density information; and a binarization process of the density information obtained by the HSV-density conversion unit to output a binary image. An optical character reading device, comprising: an image binarization unit; and wherein the HSV-density conversion unit includes a unit for rewriting conversion information for determining the content of the conversion operation.

[0029]

Embodiments of the present invention will be described below in detail with reference to the drawings. For embodiments of the present invention,
FIG. 2 is a block diagram showing the principle. The image information conversion means 2 reads a character entry frame (not only the frame itself, but also one in which the inside of the frame is colored or includes a pre-filled 8 character) and a form 1 in which the characters are entered in different colors.
It outputs image signals of three colors of RGB. RGB signal to HS
RGB-HSV conversion means 3 is provided to convert the signal into a V color system signal, and outputs a hue (H), saturation (S), and luminance (V) signal. The configuration up to this point is the same as the conventional example shown in FIG.

The output HSV signal is processed by the HSV-density conversion means 4 and output as density information. The density information is input to the binarizing means 5, and outputs a binarized signal as an output 1. Here, the HSV-density conversion means determines input / output conversion by a function having three variables of hue h, saturation s, and luminance v as inputs, and the conversion information is appropriately converted. Can be specified from outside.

A plurality of binarizing units 6 (6a, 6b,..., 6) comprising the HSV-density converting unit 4 and the image binarizing unit 5
n) Circuits are arranged in parallel. For example, each of the binarization units 6a, 6b,... 6n takes out a black character having a reddish dropout, or takes out only a reddish entry frame,
The conversion function can be specified so that only blue characters on copy paper can be extracted or can be extracted.

By appropriately selecting the conversion function as described above, even if the dropout color is made dark (conventionally, the dropout color has been required to be light), the dark color can be removed. Only specific colors can be extracted.

FIG. 4 is a structural block diagram showing an embodiment of the present invention. On the form 1, character entry frames and characters are entered in different colors. Lens 11 and color CCD
The image information on the form 1 is read line by line as image signals for three colors of R, G, and B by an optical system including the sensor 12. In this embodiment, a color CCD sensor incorporating filters for three colors of R, G, and B is used. However, the same applies to a case where three colors are read by separate CCD sensors using three optical systems. .

The image signal read by the color CCD sensor 12 is amplified to a predetermined level by amplifiers 13a, 13b and 13c, and then amplified by an A / D converter 14a.
The data is converted into digital data by 14b and 14c. Since the image signal at this point includes an offset, a shading characteristic, and the like, the normalization circuit 15 corrects these. In the present embodiment, the normalization circuit 15
It is assumed that the R, G, and B signals at the time of passing are digital values of 7 bits each.

The normalized RGB signal is represented by RGB-HS
The V conversion circuit 3 performs an arithmetic process corresponding to Equations 1, 2, and 3, and converts the data into hue (H), saturation (S), and luminance (V). In the present embodiment, in order to reduce the capacity of the filter memory 16 connected thereafter, the number of bits is reduced at the time of conversion to the HSV signal. Considering the importance of each signal, the hue H is output with 6 bits, the saturation S is output with 5 bits, and the luminance V is output with 7 bits.

The converted HSV signal is input to the filter memory 16. In the present embodiment, the filter memory 1
6 uses an S-RAM. An input signal of a total of 18 bits of HSV is connected as an address of the S-RAM, and a data signal is output as 4-bit density information (d). In this case, (2 to the 18th power) × 4 = 1M bits of S−
RAM can be applied. By changing the contents stored in the filter memory 16, an arbitrary conversion function represented by the following equation can be realized.

[0037]

(Equation 4)

First, an arbitrary function formula is applied to Equation 4, and
The output d when h, s, and v are changed is obtained in advance. In the case of this embodiment, h is from 0 to 63, and s is 0.
, And v are changed in the range of 0 to 127, and the information of the output (the range of 0 to 15, ie, 16 gradations) is stored in a table format. Hereinafter, this information is called a conversion function table. In the case of the present embodiment, the size of the conversion function table is 1 Mbit, that is, 128 Kbytes. By writing this conversion function table into the filter memory 16 by the filter memory writing circuit 18, an arbitrary filter characteristic can be obtained.

In this embodiment, the HSV-density conversion means 4 is constituted by an S-RAM, but is not limited to this, and may be constituted by another memory such as a rewritable flash memory or a D-RAM. Alternatively, the conversion function to be applied is limited and configured by an operation circuit of the conversion function (an operation circuit having a plurality of conversion functions is prepared in advance and an appropriate operation circuit is selectively adopted), or configured by a DSP or the like. And other means.

The density d output from the filter memory 16 is converted by the binarization circuit 17 into a binary image. Here, similarly to the normal binarization processing, the binarization circuit 17 includes means for improving image quality, for example, means for adaptive binarization for changing a threshold value according to the density around a target pixel. Can be applied. In addition, a means for obtaining a pseudo multi-gradation such as a known dither method or error diffusion can be applied.

Here, the filter memory 16 and the binarization circuit 17 have a plurality of systems. Filter memory 16 of each system
By changing the conversion function table described above, it is possible to obtain a different binary image for each system. For example, by giving a characteristic of dropping out reddish to the first system filter memory, an image for character recognition obtained by dropping out the character entry frame and removing only the reddish system to the second system filter memory Can be used to obtain an image in which only the character entry frame is extracted, and similarly, an image in which only the blue entry is extracted in the third system. Since the filter memory 16 is rewritable, the combination of images obtained in each system can be arbitrarily changed according to the target form or business.

The binary image obtained in each system is output by the image output means 19. In this embodiment, in addition to the binary image, the color image output from the normalization circuit 15 can be output. This is because various usages of the character recognition image are considered by outputting the image for character recognition in combination with any one of RGB colors.

A conversion function for creating a conversion function table is a function that inputs three variables of h, s, and v.
Anything can be applied. Some examples are shown in the following equations. In these equations, each parameter has a normalized value in the range of 0 to 1 (however, the hue h is in the range of 0 to 360).

[0044]

(Equation 5)

[0045]

(Equation 6)

[0046]

(Equation 7)

Equation 5 is a conversion function corresponding to a normal dropout. In the formula, fh, ws, wh are characteristic parameters, fh is the center of the hue to be dropped out, ws
Represents sensitivity to saturation, and wh represents sensitivity to hue difference. | (H-fh) / 180 | indicates the hue difference between hue h and parameter fh in the range of 0 to 1. According to Equation 5, the coefficient applied to the luminance v decreases as the hue difference from fh increases, and decreases as the saturation s increases.

If s = 0, that is, if it is an achromatic color, the coefficient applied to the luminance v is always 1, so that the achromatic characters and the background (white) are output as they are. If s> 0, that is, a chromatic color and the hue difference is not 0, that is, a color that is not a dropout color, the coefficient on the luminance v decreases as the hue difference increases. As a result, the color specified by the parameter fh passes through the luminance v as it is, so that it becomes white (background) when binarized and is dropped out.

In the conventional drop-out method, black remains black regardless of any filter. However, by forming a conversion function using luminance v as one of the variables as shown in Equation 6, the degree of freedom is extremely high. Become higher. Equation 6 is a conversion function for extracting only the designated color. In Equation 6, the specified hue f
h, the saturation s is large, and the luminance v is large, the output is small. In other words, the closer the color is to the specified color, the darker it is output. As a result of the binarization process, only the specified color is extracted.

In the conventional drop-out method, no matter how high the saturation is, if the luminance is low, the result of passing through the filter becomes dark. Therefore, there is a problem that the dropout color must be a somewhat light (bright) color. On the other hand, Equation 7 shows the color of the specified hue,
This is a conversion function that outputs lighter than other colors. The gamma value of the gamma conversion is changed according to the proximity of the designated color.
By using this conversion function, a dropout color that is darker and easier to see than before can be used.

In this embodiment, as shown in FIG.
Although the B-HSV conversion circuit 3 and the filter memory 16 are configured separately, the filter memory 16 is expanded to
The B-HSV conversion circuit 3 can be taken in. In this case, the input of the filter memory is an RGB signal, and when calculating the conversion function table to be stored in the filter memory, the RGB signal is used.
The calculation including the B-HSV conversion may be performed.

FIG. 5 is a system configuration diagram showing an embodiment of the present invention. The scanner device 27 includes a control unit 24 that controls the mechanical system and the entire scanner device, an interface control unit 23 that controls the interface 22 with the host device,
It comprises a reading unit 26 corresponding to FIG. 4 and an image memory 25 for temporarily storing read image information. The host device is connected to the scanner device 27 via the interface 22, and performs a character recognition. The recognition unit 20 controls the scanner device 27 via the interface 22, processes image data transferred from the scanner device, and executes a user interface. Software such as an application 21 for performing processing such as the above is operating.

Hereinafter, a method of writing the conversion function table will be described with reference to FIG.

In the first method, a conversion function table is obtained in advance and stored in a host device. Or,
Then, before starting the reading operation, the conversion function table is transferred from the host device to the scanner device 27, and the reading unit 2
6 is written into a predetermined filter memory 16. Alternatively, a standard conversion function table is stored in the scanner device 27 and written in a predetermined filter memory 16 before the reading operation.

According to the second method, a conversion function parameter is specified as one item of format information in an application for creating format information specifying a reading position of a form to be read. Here, the conversion function parameter is a type of the conversion function to be applied and a parameter given to the conversion function. Then, before starting reading, the host device performs a calculation process based on the conversion function parameter to create a conversion function table, transfers the table to the scanner device 27, and writes it to the predetermined filter memory 16.

In the third method, first, one form is read and transferred as a color image. Here, since the color image has a larger data amount than the binary image, the processing time should be longer. Then, the operator specifies a conversion function parameter for the captured color image. At this time, it is assumed that parameters equivalent to the reading unit 26 in FIG. 5 can be interactively specified by performing arithmetic processing on the host device and displaying a preview screen on the screen.

A conversion function table is created by arithmetic processing based on the determined conversion function parameters, transferred to the scanner device 27, and written into the filter memory 16. As a result, since the second and subsequent sheets perform the binarization processing on the scanner device 27 side, high-speed operation becomes possible. Further, the created conversion function table can be stored in a recording medium, and the conversion function table stored later can be designated and reused, or can be associated with the format information.

FIG. 6 shows an example of a conversion function parameter setting screen. That is, a conversion function used in the filter of FIG. 4 is selected on the screen, parameters to be given to the conversion function are appropriately set on the screen, and colors are specified to simulate how the filter output will be. This is an example of how to show them.

In the filter 28, the type of the conversion function to be applied is selected (any of the conversion functions shown in Expressions 5, 6, and 7 is selected). The color difference sensitivity 29 (wh in Equation 5, 6, or 7) and the luminance sensitivity 30 are parameters given to the conversion function determined in the item of the filter 28. In the color designation 31, hue is given on the color chart.
When these parameters are given, the same calculation processing as that of the reading unit 26 of the scanner device 27 is performed, and a chart of density output and binarized output is displayed. For example, if the color designation 31 is magenta as the color to be dropped out, the filter output 32 is dropped out around magenta and becomes the same as the background, so that the binary image 33 can be previewed. A similar preview may be performed on a form image captured as a color image from the scanner unit 27.

Here, since the transformation function takes three variables of h, s, and v as inputs, it is difficult to completely graph the characteristics and display them on the screen.
Must match the characteristics of the transformation function.

As described above, the embodiments of the present invention include those having the following configurations and functions. That is, without physically exchanging the filters, a plurality of filters in which the filter characteristics for the converted density signal are appropriately changed are used to obtain a binarized image, so that a dark dropout color or a specific color is obtained. It is possible to obtain a binary image of extraction, reading of colored characters, and the like. Since the filter characteristics can be appropriately changed, an optical character reader having a high degree of freedom in the filter characteristics can be obtained.

[0062]

The optical character reading apparatus of the present invention converts an RGB signal obtained from a form into an HSV signal, and further comprises a rewritable HSV-density converting means, so that the filter is not physically replaced. Thus, a binarized image with changed filter characteristics can be obtained.

The input of the HSV-concentration conversion means is h,
By using the three variables s and v, an arbitrary conversion function can be realized, and it is possible to flexibly cope with a form that has been difficult to handle conventionally, such as removal of a dark dropout color and extraction of a specific color. it can.

Further, by providing a plurality of systems after the HSV-density conversion means, it is possible to obtain a binary image for a plurality of purposes by one reading operation.

Further, since the color filter processing is provided by hardware, conversion processing is performed simultaneously with reading, and a binary image is output, the performance is not reduced as compared with the conventional system.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating an HSV color system.

FIG. 2 is a principle block diagram of the present invention.

FIG. 3 is a block diagram showing the principle of the technique described in Japanese Patent Application Laid-Open No. 2-67689, which is a conventional example.

FIG. 4 is a configuration block diagram illustrating an embodiment of the present invention.

FIG. 5 is a system configuration diagram showing an embodiment of the present invention.

FIG. 6 is a diagram showing an example of a conversion function parameter setting screen.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Form 2 Image information conversion means 3 RGB-HSV conversion means 4 HSV-density conversion means 5 Image binarization means 6 Binarization part 7 Conversion information 8 Achromatic / chromatic color judgment means 9 Background / character judgment means 10 Color judgment Means 11 Lens 12 Color CCD sensor 13 Amplifier 14 A / D converter 15 Normalization circuit 16 Filter memory 17 Binarization circuit 18 Filter memory writing circuit 19 Image output means 20 Recognition unit 21 Application 22 Interface 23 Interface control unit 24 Control Unit 25 image memory 26 reading unit 27 scanner device 28 filter setting column 29 color difference sensitivity setting column 30 luminance sensitivity setting column 31 color designation color chart 32 filter output preview image 33 binarized output preview image

──────────────────────────────────────────────────続 き Continued on front page (51) Int.Cl. ⁶ Identification code FI H04N 1/46 H04N 1/46 Z

Claims (5)

[Claims] 1. An image information of a form is read and RGB is read.
And converting the RGB signals into a hue (H) and a saturation (S)
And a signal (HSV signal) relating to the luminance (V), and the HSV signal is input to a plurality of HSV-density conversion means having respective conversion function tables to be converted into respective density signals. The signal is binarized and converted into a plurality of systems of binary images, and for each system, respective binary images for different reading purposes including an image for character recognition by removing a dropout color and an extracted image of a specific color are obtained. An optical character reading method, comprising: 2. The optical character reading method according to claim 1, wherein the conversion function table is composed of types of conversion functions to be applied and parameters given to the conversion functions, and can be arbitrarily created. Characteristic optical character reading method. 3. An image information converting means for reading a form as an RGB image signal;
RGB-HSV conversion means for converting the GB signal into hue (H), saturation (S), and luminance (V); and at least one for converting the HSV signal obtained by the RGB-HSV conversion means into a binary image. A binarizing unit, wherein the binarizing unit is obtained by an HSV-density converting unit that converts the HSV signal obtained by the RGB-HSV converting unit into density information, and obtained by the HSV-density converting unit. An image binarization unit that binarizes density information and outputs a binary image, wherein the HSV-density conversion unit includes a unit that rewrites conversion information that determines the content of the conversion operation. Optical character reader. 4. The optical character reading device according to claim 3, wherein the HSV-density conversion means includes a rewritable memory;
Or an arithmetic circuit or DSP embodying the conversion information
An optical character reading device comprising: 5. The optical character reading device according to claim 3, further comprising: a unit that obtains a simulated image of a binary image obtained by the image binarizing unit, wherein the converting operation of the HSV-density converting unit is performed. An optical character reading apparatus characterized in that a type of a conversion function to be applied and parameters to be given to the conversion function are arbitrarily set as conversion information for determining, and a simulation image with each setting is obtained.