JPH05501778A - Method and apparatus for automatic text separation using automatic electronic filtering of multiple dropout colors for optical character recognition of preprinted forms - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] 11 Dropout car filter The ring was for... Departure Ω technical minute! This invention provides automatic selection and detection of dropout colors using a color electronic scanner. More specifically, optical character recognition (COR) systems Automatically filter based on format itself instead of matching optical filters allows you to adjust parameters.

amount of skin growth Optical character recognition (OCR) is an effective technique for processing office forms.

Machine-reading systems replace several data entry operators and reduce the cost of data capture. It is possible to reduce the amount of use.

-C, the first step in the OCR process is to scan the document electronically to extract all of the information. It is to convert to digital bit manob. - The image is captured in electronic form. The information to be read is then separated from the background information. No box frame or guide text The text of the entry should be read. This separation is complete. Once created, the electronic image of the text is processed by an OCR algorithm, during which Characters of interest are converted to ASCII data.

Almost all OCR systems that process office forms use "dropout color". -' technique is used. Print documents in predetermined colors (usually pastel colors) and documents by using color-matched optical filters in an electronic scanner. The text can be separated from the print format and the 0-color filter can be scanned cause the device to ignore information printed in that color (for W reservation scanners, the color of the form is (appears equivalent to the white background of the paper), while the text This information is typically typed or printed in black (or other dark colors). The information is captured by a scanner. Therefore, the preprinted format is on a white background. The input text is therefore easily processed by OCR algorithms. be able to.

The use of optical filters works well in this application, but it very specific colors (precisely due to the characteristics of the optical filter attached to the scanner) (consistent with). by adding additional optical filters. Additional dropout colors can be included in the scanner. Therefore, certain books Form processing involves selecting a suitable optical filter and mechanically processing the form before processing it. will need to be inserted into.

However, slight variations in the printing process may cause variations in the actual color of the printed form. This may reduce the "dropout" effect. the Such changes can add noise (scanners do not print pre-printed formatting information in white). (see as black), which causes the OCR algorithm to produce erroneous results. 6 Alternatively, the light can be adjusted to accommodate these slight variations during printing. Although it is not practical to change the analytic filter, it is important to note that this change would require multiple filters with slightly different characteristics. It's your body. Therefore, the only way to practically adjust this problem is to modify the printing process. The goal is to ensure uniform dropout color through strict control. The result and Therefore, OCR format reading systems currently in use are generally “closed loop”. This is because formats produced by outside companies may not read properly due to color variations. The form processing company (e.g. insurance company) may not be able to assist you in printing the form. It means you have to maintain control.

In this invention, the scanner can separate all images into three primary colors: red, green and blue. Probably. A black and white representation of an image can be created simply by adding the two color components. I can do it. By processing the red, green and blue signals independently, the device Filter the colors to keep only high-contrast text for OCR reading. Uni, common white! Color information can be separated from information.

Dispatch■Return In this invention, to ensure uniform color and amplitude response among all pixels, The three digital channels are multiplied by appropriate coefficients. These three signals , - once corrected for uniformity, is processed as an independent video signal and displayed in the image. Generate three binary representations of the image. This signal is common to all three channels. By combining in such a way that only the information of A "full color" filter is generated that can separate "black" text from print information. Ru. Effectively, the two outputs represent all possible dropout colors in the color space. A document image is expressed using a combination of

Ward episode Ω solid fist preaching place Figure 1 illustrates a solid-state charge-coupled device configuration that can be used for color scanning. , FIG. 2 shows the structure of the circuit used for electronic color filtering according to the invention. The composition diagram is illustrated, and 3A-B illustrate a flowchart used in connection with black and white proofing.

1 treatment 1 method for skin regrowth Figure 1 shows the electronics used to generate programmable dropout colors. A typical example of a digital scanner is illustrated. This scanner scans all images in three primary colors, red , green, and blue (a typical electronic scanner (such as the one we generate today) is generated by simply adding the three color components be able to.

The electronic scanner intended for use in the device of this invention is ``Contact'' currently available as model TCD126C made by It is based on the JCCD (charge coupled device) 10. This COD is actually a single several CCD arrays on a substrate with a horizontal resolution of 1200 pixels/pixel. It has a length of 12 inches. Most OCR algorithms are less than 200 400 pixels/inch scanning resolution allows for accurate reading Additional resolution can be used for detection. Such detection is This is done by covering adjacent pixels with appropriate red, green and blue optical filters. , the spectral content of these filters is based on the spectral characteristics of the CCD element itself. I'm there.

As shown in FIG. 1, three adjacent cells 12, 14, and 16 are connected to a single r 1 pixel" 18, red, green and blue optical filters 20, red respectively, 26, green 28, and blue 30 video signals.

Figure 2 shows automatic text separation for OCR reading and for use in full image capture. This diagram shows the first blog diagram. Color scanner 10 has three video signals per pixel. - 26 red, 28 green, and 30 blue colors are output in a segmented manner for each scan line. child These R, G, and B signals are grouped by respective A/D converters 32, 34, and 36. Converted to a race-scale digital representation. The red, green, and blue components of each pixel are are fed to multipliers 38, 40 and 42, respectively.

A microprocessor and RAM storage subsystem 52 monitors each pixel within the scan line. to ensure proper correlation between pixel video data and calibration coefficients 38, 40 and 42. and these coefficients are applied to the corresponding multipliers 46.4 for their respective color channels. 8 and 50. The output of these multipliers is a grayscale color image and Calibrated Red 56, Green 57, and Blue 58i! ii element segmentation pi It has a modern style. This calibration color information is also supplied to the gilding joint 59, Here the three color components for each pixel are added and the output is a grayscale form a black and white image. In addition, calibration red 56, green 57, and blue 58 video data is provided to the microprocessor/RAM storage subsystem 52 for diagnostic purposes. It will be done.

The calibrated red 56, green 57, and blue 58 video data is also )/N element video data generating threshold circuits 41, 43 and 45, respectively; Processed by Threshold circuits 41, 43 and 45 take the form of simple comparators. or like an MXN convolution filter with adaptive threshold settings. The output of each threshold circuit 41, 43 and 45, which may be sophisticated, is a binary value. , rl corresponds to a "dark" pixel, and "0" corresponds to a "bright" pixel.

All three of these binary signals (corresponding to red, green and previous value for each pixel) ) is sent to AND gate 63. Which of the three color components of a given pixel is "light" or "0", the AND gate corresponds to "white" and returns "0". If all three color components of the 0 pixel are "1", that is, "dark", then A The output of the ND gate becomes "1" corresponding to "black".

The output of AND gate 63 is a typical book using pastel dropout colors. The 0-color background information that could be considered the ``body output'' for the expression is removed and typed. Only the text information that has been detected is sent to the OCR algorithm. For example, non-carbon red ink typewriter using carbon ribbon Equations could be easily processed using this invention. All printing fees Normal scanning to produce an image of the material (pre-printed red and typed genera) system, the system of this invention ignores the pre-printed red and will only generate an image of the typed text.

Optical red filter closely matches the ink color used for pre-printed information 1 (below) by using this filter and with a conventional scanning system. able to achieve results. However, optical red filters do not drive green ink. Will not be able to drop out. Advantageously, this invention Carbon ink can be removed, thereby giving greater flexibility .

Users can use mixed-color documents without worrying about changing filters. will be possible.

Therefore, any dropout color used in a particular format will detected automatically by making certain assumptions about the spectral content of the text. and can be suppressed. most used for filling out office forms The main text is ``Carbon can be classified as principal component formula J. Includes standard typing ribbon, pen or pencil. Such a text was used Since it will work regardless of any color filter, the main text is ``all color filters''. Any pre-printed color information can be separated from Ru.

car color calibration White calibration is a pixel-by-pixel method to compensate for any spectral anomalies or sensitivity variations. can be used to optimize scanner performance. discussed here In the white calibration method, the compensation is done immediately before operation, which prevents aging or TIJ losses. To ensure a uniform response from the scanner, This is the preferred method.

When a white (blank) piece of paper is fed through a color scanner, all three color signals are detected. work at the same time. Since a piece of white paper has a known and predictable spectral curve, The color gain factor is programmed in such a way as to allow the scanner to mimic this ideal response. can be programmed. Figures 3A and B are flowcharts for performing white calibration. It shows. Step 80 sets all of the red, green and blue gain factors to 1. microprocessor-RAM storage subsystem 52 (FIG. 2) to set the value to and then in step 82 (microprocessor RAM storage subsystem All of the pixel accumulators (located in system 52) are set to O. In step 84 , the operator feeds a piece of white paper through the color scanner to calibrate the response. step At floor 86, the beginning of the page is detected and the proofing process begins. color scanner 10 When scanning each horizontal line of a blank document, sequential three-color data 'f, (R, G, B) is output. This information is transmitted to the A/D converter 3, one for each chaffle. 2.34 and 36. The digitized signals are each multiplied by It is sent to calculators 38, 40 and 42. Microblosse, Su 52 has all the benefits before Since the value is set to 1, the output of each multiplier is equal to the R, G, and B values of each pixel. It is worth it. The microprocessor 52 converts the grayscale color information of this sequential line into stages. capture into its own memory (RAM) at step 88 and then step 90 Add the red, green and blue values of each pixel to the appropriate accumulators accordingly.

The microprocessor has separate accumulators for R, G, and B values for each pixel. (total number of accumulators = 3 x number of horizontal pixels). This accumulation process The process continues until the end is detected at step 92. The total number of lines processed is in stages 94 by a line counter. Once the page has been scanned, the microphone The processor 52 divides each accumulator value by the line count (number of lines captured). Thus, in step 96, the average red, green, and previous values for each pixel are calculated. this The information corresponds to the average color response for each horizontal pixel.

Once this color response is known, the red, green, and blue coefficients are determined for each pixel in step 98. can be calculated as follows. This is done to "normalize" the response, but this Normalization means that each pixel responds in a similar manner when given inputs of similar type A or higher. Guarantee. The gain coefficient is the ideal or optimal R, G, B response for each pixel. Calculated by dividing by the B response. Optimal response is ideal for white human power R.

Based on G and B values. The gain factor is 3 per pixel, calculated in steps 100 A dual-port storage device (not shown, but a microprocessor, A part of the AM memory subsystem) and microprocessor/RAM memory. Subsystem 52 thereby completes the white calibration process. - Once calibrated , the device (Figure 2) multiplies the red, green and blue video values of each pixel by an image compensation factor. Any color or gain anomaly can be compensated for by In operation, color scanner The IO sequentially outputs red, green and blue signals for each horizontal pixel, and each color signal are digitized by A/D converters 32, 34 and 36. for each pixel The digital grayscale color information is then applied to multiplier circuits 38, 40 and 4, respectively. Sent to 2. The microprocessor/RAM storage subsystem 52 is configured for horizontal scanning. Call the unique RlG,B gain coefficients for each pixel in The red, green, and previous values of each pixel are applied to three multipliers simultaneously, thereby converting the red, green, and previous values of each pixel into their corresponding Multiply by the gain factor. The output of these multipliers is the normalized red, green for each pixel. and zero calibration representing the previous value and memorize the unique color gain coefficient for each pixel. , by then normalizing the R, G, B response for each pixel using a gain factor. Therefore, the output of color scanner 10 is balanced for the correct uniform spectral response. It will be done.

Although this invention has been described in connection with particular embodiments thereof, many alternatives, modifications, and and variations thereof will be apparent to those skilled in the art in light of the foregoing description. be. Therefore, all alternatives that come within the spirit and broad scope of the appended claims. Examples, modifications, and variations are intended to be covered.

Fortune-telling This invention separates text information in a format by automatically filtering color information. Useful for processing office forms in conjunction with optical character recognition systems as a method of It is effective. This scanner system scans all images in three primary colors: red, green and The black and white representation of the 0 image, which would be decomposed into blue and blue, can be achieved by simply adding three color components. can be generated. The present invention provides a droplet associated with mechanical filter insertion. Advantageous in eliminating pull-out color capability problems. This variability is used in formatting. This can be caused by the color of the ink changing from printing batch to printing batch. , mechanical filters remove printed text on forms printed with non-acceptable inks. He was incompetent to leave.

Moreover, the present invention allows for the mixing of different colored documents within a batch and the ability to Documents with different dropout colors (e.g. red and blue pre-printed) (format that prints information). Without the use of this invention, the prior art It is not possible to do this using mechanical filter insertion as was done in Probably.

FfG, 3B Automatic electronics with multiple drop-out colors for optical character recognition of pre-printed forms. Method and apparatus for automatic text separation using child filtering ! -Target-1 The text used to fill out the office form was classified as ``Carbon Principal Component Formula J''. and such a body can be used regardless of any color filter used. Separate text from pre-printed material based on the assumption that it will pass as black. method and apparatus for The main text can be changed to any color by applying the "all color filter". Can be separated from preprint color information.

International search report. rTVl,. 017ncnA111-To hit Uni? Hit--- ------1 to 2? 22 o----old 2 nee-5oo6 艷2----ichimon 2 Tapping two gloss---□ □

Claims (11)

[Claims] 1. A three-component color signal generated by a color scanner is defined by the signal components as follows: Graying is performed on a pixel-by-pixel basis by an analog-to-digital converter associated with each color component. - A device for processing after conversion to scale digital format, which Three binary values, one bit per pixel bit, each corresponding to the red, green and blue components of the pixel. a processing device for each of the color video signals to produce a video signal; to Stores black information only when all of the binary color signals identify a pixel as black a device for combining said color video signals, characterized by An apparatus for processing said three-component color signal. 2. 2. The device of claim 1, wherein the combination device is in the form of an AND gate. 3. The output of the combination device described above is in one bit per pixel section suitable for OCR reading. 3. The device of claim 2, which is a differentiated scan line. 4. Each pixel in a scan line has an amplitude and color response as represented by three color components. 2. The device of claim 1, further comprising a device for compensating for. 5. To process color formats written using carbon-based inks A device, Multiple grays that generate scan lines with segmented pixels with red, green, and blue components A color scanner with scale color outputs, each output grayscaled in a pixel-by-pixel manner. equipment for converting files into digital format; a storage device for storing at least one scan line of color grayscale information; , For amplitude and color response such that each pixel in a scan line is represented by three color components equipment for compensating; Processes each of the color video signals to separate the red, edge and blue components of the pixel. In order to generate a corresponding three binary value 1 bit per pixel video signal equipment, and While all of the binary color outputs above are displaying the pixel as being black, Binary color to preserve the black information associated with carbon-based ink devices for combining information; Apparatus for processing said color format, characterized by: 6. 6. Apparatus according to claim 5, wherein the compensation device is in the form of a digital multiplier. 7. How to process color formatting written using carbon-based inks And, Scan the color format to create at least two segmented pixels for each color component. The stage of generating grayscale color output, grading each output in a pixel-by-pixel manner. converting to scale digital format, at least one of each grayscale color component; A step in which each pixel in a scan line is stored as represented by each color component. compensating for amplitude and color response; Each color video signal is processed so that each color component corresponds to each pixel. generating three corresponding binary bit-per-pixel video signals; Comparing the binary color information of, all of the color components display the pixel as black. a step of preserving black information associated with carbon-based ink; Said method characterized by. 8. the combining step is meant to AND all of said outputs; A method for processing color formats according to claim 7. 9. Processing a color format according to claim 7, wherein the combined output is suitable for OCR reading. How to manage. 10. The sum of the grayscale color outputs produces a grayscale black and white image. A method for processing color formatting as described in claim 7. 11. the grayscale color output signal produces a grayscale color image; A method for processing color formats according to claim 7.