CN104346609A - Method and device for recognizing characters on printed products - Google Patents
Method and device for recognizing characters on printed products Download PDFInfo
- Publication number
- CN104346609A CN104346609A CN201310331468.4A CN201310331468A CN104346609A CN 104346609 A CN104346609 A CN 104346609A CN 201310331468 A CN201310331468 A CN 201310331468A CN 104346609 A CN104346609 A CN 104346609A
- Authority
- CN
- China
- Prior art keywords
- image
- gray
- scale value
- module
- character
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/22—Image preprocessing by selection of a specific region containing or referencing a pattern; Locating or processing of specific regions to guide the detection or recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
- G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Character Input (AREA)
Abstract
The invention relates to a method and a device for recognizing characters on printed products. The method can comprise the steps of shooting the printed product for obtaining a to-be-recognized image; copying the image for obtaining at least two copied images, and respectively carrying out different image processing on each copied image for obtaining at least two layered images; carrying out layer merging on the obtained layered images for obtaining a processed image; extracting an image of each character from the processed image; carrying out character recognizing on the extracted image of each character. By adopting the technical scheme disclosed by the invention, when image processing is carried out on the printed products such as credentials, more effective and more accurate recognizing on the characters on the printed products can be realized.
Description
Technical field
The application relates to image identification technical field, particularly relates to a kind of method and the device that identify character on printed matter.
Background technology
At OCR(Optical Character Recognition in the past, optical character identification) identify in, for some outside smoother, reflect the identification of word on stronger printed matter, the identification of the word such as on the printed matter of surface coating process, or such as all kinds of certificate photo or various card (especially cross the certificate (driver's license moulding process, driving license etc.)) identification of upper word, often there is discrimination lower or in addition because surface coating is reflective thus there is the situation of identification error, the essence of this problem is caused to be cannot effectively filter in the process identified, causing OCR to identify there is fuzzy or that contrast is excessive problem in the font in source, simultaneously, because various printed matter often exists multiple different font, also the character that bring cannot be able to mate or the problem of matching error on identifying.
At present, in the recognition technology development of OCR, demand towards license gets more and more, and the developing direction of existing OCR technology all trends towards identification for complete image information and search, for current license identification, in existing several OCR identifying schemes, for I.D., the identification of passport etc., although there are comparatively ripe high discrimination engine and algorithm present stage, but for similar driving license, in the identification of employee's card etc., because these certificates all can carry out moulding process to certificate when finally issuing, simultaneously also because the similar certificate in each area prints and possess unified printing standard and font unlike I.D., thus result in existing license identification, for need the license identified often exist due to over-exposed cause image blurring and for the low problem of recognition efficiency of distortion font, say from essence, be the needs not considering these two aspects in existing recognition methods completely.
Summary of the invention
The fundamental purpose of the application is to provide a kind of method and the device that identify character on printed matter, with solve prior art exist character on printed matter is identified in image processing problem and character recognition problem, wherein:
According to an aspect of the application, provide a kind of method identifying character on printed matter, it is characterized in that, comprising: take described printed matter to obtain the image that will identify; Copy to obtain at least two width duplicating images to described image, and different image procossing is carried out respectively to obtain at least two width layered images to every width duplicating image; The layered image obtained is carried out layer merging, to obtain processing rear image; The image of each character is extracted image after described process; And character recognition is carried out to the image of each character extracted.
According to the embodiment of the application, in the method, take described printed matter to obtain the image that will identify, comprising: carry out exposure when taking by predetermined condition and arrange.
According to the embodiment of the application, in the method, different image procossing is carried out respectively to obtain at least two width layered images to each width duplicating image described, comprising: the process of removal noise is carried out to obtain first layer image to the width in described duplicating image; And contrast enhancement processing is carried out to obtain the second layered image to another width in described duplicating image.
According to the embodiment of the application, in the method, the process of removal noise is carried out to obtain first layer image to the width in described duplicating image, comprising: identify the noise in described duplicating image; Using the gray-scale value of each noise with its around the gray-scale value phase adduction of adjacent eight pixels average as the denoising gray-scale value of each noise; And the denoising gray-scale value gray-scale value of noise each in described duplicating image being replaced with this noise is to obtain first layer image.
According to the embodiment of the application, in the method, identify that the noise in described duplicating image comprises: using the gray-scale value of pixel each in described duplicating image with its about the gray-scale value phase adduction of two neighbor pixels average as the calculating gray-scale value of each pixel; Judge that the absolute value of the gray-scale value of each pixel and the difference of its calculating gray-scale value is whether in predetermined threshold range; And the pixel that gray-scale value and the absolute value of the difference calculating gray-scale value exceed predetermined threshold range is identified as noise.
According to the embodiment of the application, in the method, contrast enhancement processing is carried out to obtain the second layered image to another width in described duplicating image, comprising: described duplicating image is divided at least two sub regions; And respectively gray scale adjustment is carried out to every sub regions, to obtain the second layered image.
According to the embodiment of the application, in the method, described layered image is merged, to obtain processing rear image, comprising: intermediate value is got to the gray-scale value of pixel corresponding in described layered image, obtain the gray-scale value intermediate value of each pixel; And the gray-scale value of each pixel is replaced with the gray-scale value intermediate value of this pixel, to obtain processing rear image.
According to the embodiment of the application, in the method, extract the image of each character in image after described process, comprising: the position determining the text image after described process in image; And Character segmentation is carried out to described text image, extract the image of each character in described text image.
According to the embodiment of the application, in the method, obtain the position of the text image after described process in image, comprising: identify the Edge texture in every row pixel by rim detection; Histogram is done to the Edge texture of every row pixel, and according to described histogrammic analysis being determined to the recognition threshold of edge primitive; According to the number of the edge primitive that the recognition threshold statistics of described edge primitive is often gone, and record starting position and the end position of often row edge primitive; Identify the non-blank-white row in image after described process; Judge whether current non-blank-white row meets pre-conditioned, if met, then carry out the detection of next non-blank-white row; And when be consecutively detected the non-blank-white row exceeding predetermined number meet described pre-conditioned time, according to the starting position of each non-blank-white row edge primitive and the position of end position determination text image.
According to the embodiment of the application, in the method, character recognition is carried out to the image of each character extracted, comprising: utilize the image of BP neural network to described each character to carry out character recognition.
The another aspect of the application, provides a kind of device identifying character on printed matter, it is characterized in that, comprising: acquisition module, obtains the image that will identify for taking described printed matter; Hierarchical processing module, obtains at least two width duplicating images for copying described image, and carries out different image procossing respectively to obtain at least two width layered images to every width duplicating image; Layer merges module, carries out layer merging, to obtain processing rear image for the layered image that will obtain; Extraction module, for extracting the image of each character in image after described process; And identification module, for carrying out character recognition to the image of each character extracted.
Compared with prior art, according to the technical scheme of the application, by taking printed matter and carrying out layered image process to the image that will identify, and carry out effects compensate by layer merging, can picture quality be promoted, improve the accuracy rate identified.
Accompanying drawing explanation
Accompanying drawing described herein is used to provide further understanding of the present application, and form a application's part, the schematic description and description of the application, for explaining the application, does not form the improper restriction to the application.In the accompanying drawings:
Fig. 1 is a kind of process flow diagram identifying the method for character on printed matter of the embodiment of the present application;
Fig. 2 is in the step S102 in Fig. 1 of the embodiment of the present application, removes the process flow diagram of the step S1 of noise process;
Fig. 3 is the process flow diagram of the step S201 in Fig. 2 of the embodiment of the present application;
Fig. 4 is in the step S102 in Fig. 1 of the embodiment of the present application, the process flow diagram of the step S2 of contrast enhancement processing;
Fig. 5 is the process flow diagram of the step S103 in Fig. 1 of the embodiment of the present application;
Fig. 6 is the process flow diagram of the step S104 in Fig. 1 of the embodiment of the present application;
Fig. 7 is the process flow diagram of the step S601 in Fig. 6 of the embodiment of the present application; And
Fig. 8 is a kind of structured flowchart identifying the device of character on printed matter of the embodiment of the present application.
Embodiment
The main thought of the application is, by taking the printed matter with word, become at least two width images to carry out different image procossing respectively the copying image obtained and obtain layered image, and layer merging is carried out to each layered image, obtain the image after processing, then Text Feature Extraction and Text region are carried out to the image after described process.
For making the object of the application, technical scheme and advantage clearly, below in conjunction with the application's specific embodiment and corresponding accompanying drawing, technical scheme is clearly and completely described.Obviously, described embodiment is only some embodiments of the present application, instead of whole embodiments.Based on the embodiment in the application, those of ordinary skill in the art are not making the every other embodiment obtained under creative work prerequisite, all belong to the scope of the application's protection.
According to the embodiment of the application, provide a kind of method identifying character on printed matter.
The character that the application can be applied to being printed on the printed matter of character identifies, such as, may be used for the identification of certificate, especially identifies excessively moulding the certificate processed.
A kind of method flow diagram identifying character on printed matter with reference to figure 1, Fig. 1 is the embodiment of the present application: as shown in Figure 1, in step S101, takes described printed matter to obtain the image that will identify.
When taking, because image capture device is uneven, the impact of various aspects may be subject to during shooting, such as time shutter, exposure compensating etc., the image effect taking out may be caused bad, also can affect the follow-up process to image simultaneously.Therefore, when taking, before shooting, exposure can be carried out by predetermined condition and arranging, obtaining the picture of better effects if.The situation that the optimum configurations that during by taking under equivalent environment conditions such as () such as light intensity image of the same type, exposure is relevant is different and produce different-effect is added up, and sets described predetermined condition.
In step s 102, copy to obtain at least two width duplicating images to described image, and different image procossing is carried out respectively to obtain at least two width layered images to every width duplicating image.That is, the copying image of shooting is become many parts, respectively image procossing is carried out to each the width duplicating image obtained, and be different to the image procossing that every piece image carries out, this is just equivalent to carry out layered shaping to original image, thus obtains the layered image through different disposal.
Described different image procossing can comprise: remove noise process, contrast enhancement processing.Can also comprise other image procossing, such as, path coloring treatment, pattern cut process, texture recognition pre-service etc., after these image procossing, will obtain several layered images.
Step S102 may further include step: S1 carries out the process of removal noise to the width in described duplicating image and obtains first layer image; And step S2 carries out contrast enhancement processing to another width in described duplicating image and obtains the second layered image.
Fig. 2 is the particular flow sheet carrying out the step S1 removing noise process, and as shown in Figure 2, step S1 can comprise:
Step S201, identifies the noise in described duplicating image.As shown in Figure 3, step S201 may further include sub-step S301-S303.
In sub-step S301, using the gray-scale value of pixel each in described duplicating image with its about the gray-scale value phase adduction of two neighbor pixels average as the calculating gray-scale value of each pixel.
In sub-step S302, judge that the absolute value of the gray-scale value of each pixel and the difference of its calculating gray-scale value is whether in predetermined threshold range.
In sub-step S303, the pixel that gray-scale value and the absolute value of the difference calculating gray-scale value exceed predetermined threshold range is identified as noise.Wherein predetermined threshold range can be arranged according to concrete condition, or can also arrange according to the empirical value carrying out accumulating in noise identification and processing procedure in the past.
Step S202, after identifying the noise in described duplicating image, using the gray-scale value of each noise with its around the gray-scale value phase adduction of adjacent eight pixels average as the denoising gray-scale value of each noise.Because pixel is evenly arranged with both direction in length and breadth, therefore, each pixel can have eight adjacent pixels, and therefore, the gray-scale value of eight pixels be adjacent by the gray-scale value of each noise carries out suing for peace at the denoising gray-scale value as this noise of averaging.
Step S203, obtains first layer image by the denoising gray-scale value that the gray-scale value of noise each in described duplicating image replaces with this noise.After obtaining the denoising gray-scale value of each noise, the gray-scale value of each noise in described duplicating image is replaced with the denoising gray-scale value of this noise, and the gray-scale value of other pixel (not being noise) being constant, obtaining the first layer image through removing noise process.
Digital picture, in gatherer process, due to reasons such as illumination or objects itself, often there will be the situation that target area contrast is low, can carry out contrast enhancement processing to image.
Fig. 4 is the process flow diagram of the step S2 described duplicating image being carried out to contrast process, and as shown in Figure 4, step S2 can comprise:
Step S401, is divided at least two sub regions by described duplicating image.
Carry out the basic thought of contrast enhancement processing, be by image by being divided into two sections or multistage between gray area, carry out greyscale transformation respectively, thus strengthen the contrast of image.
First, can determine by the analysis of the grey level histogram to duplicating image the boundary threshold dividing the number of subregion and the subregion of division.Grey level histogram is the pixel frequency of occurrences of different grey-scale in statistical picture, therefore the distribution situation of described duplicating image gray-scale value can be obtained according to grey level histogram, and determine image to be divided into multiple subregion according to the distribution situation of described duplicating image gray-scale value, and determine that the boundary threshold of zoning is to determine the waypoint in adjacent two regions, and by waypoint, described duplicating image is divided at least two sub regions.In the division of subregion, how many crests or trough can be had to determine to divide the number of subregion according to the grey level histogram of image, and using paddy as subzone boundaries threshold value.In the setting of boundary threshold, training can be carried out determine according to image engine, namely, the suitable boundary threshold determined is trained to the similar image of recognition image of wanting in a large number, and the determination of waypoint can calculate according to the boundary threshold chosen, or threshold value can also be set on the histogram to determine waypoint.
Step S402, carries out gray scale adjustment respectively to every sub regions, to obtain the second layered image.
Respectively gray scale adjustment is carried out to every sub regions, specifically, be exactly as required, each pixel in every sub regions is carried out the conversion of gray-scale value according to pre-defined rule, with between the gray area at Prwsnt targets of interest place, relatively suppress those uninterested gray spaces, can linear transformation be adopted, namely utilize predetermined linear transformation for mula to carry out the conversion of gray-scale value, and obtain the second layered image.
In step s 103, the layered image obtained is carried out layer merging, to obtain processing rear image.
Fig. 5 is the particular flow sheet of step S103, and as shown in Figure 5, step S103 can comprise:
Step S501, gets intermediate value to the gray-scale value of corresponding pixel in described each layered image, obtains the gray-scale value intermediate value of each pixel.
Specifically, the every width layered image obtained in above-mentioned step S102 is all carry out the image after different images process respectively for each identical duplicating image, therefore the pixel that the pixel in each width layered image is still still original, the still identical graphical information expressed, just after different image procossing, the gray scale of each pixel may create change, therefore, intermediate value is got to the gray-scale value of pixel corresponding in every width layered image, a suitable new gray-scale value can be determined for each pixel.
Step S502, replaces with the gray-scale value intermediate value of this pixel by the gray-scale value of each pixel, obtain processing rear image.
Specifically, can in the original image taking the image obtained or another width duplicating image, using the gray-scale value intermediate value of each pixel that the obtains new gray-scale value as this pixel, the gray-scale value of each pixel is adjusted to the gray-scale value intermediate value of this pixel, obtain processing rear image, the layer that this completes layered image merges, and obtains the image after process.Alternatively, after completing image processing and tracking unit, consider the needs of picture quality, the gray scale pixel conformed to a predetermined condition in image after the described process obtained can also carried out again is painted, thus more intentinonally image is marked on image, such as, 2 are added, to promote the color depth of partially black pixel by leveling off to the gray-scale value of pixel of black pixel (gray-scale value exceedes the point of certain value).
To image after the process obtained, the contrast with original image can also be carried out, the gray-scale value of each for the image after described process gray-scale value of pixel and the corresponding pixel points of original image is subtracted each other the gray value differences obtaining each pixel, and judge whether the absolute value of described gray value differences exceedes predetermined threshold, if the gray value differences of this point exceedes predetermined threshold, then also need the adjustment gray-scale value of this point being carried out to gray-scale value.
In step S104, after described process, image, extract the image of each character.
Be the particular flow sheet of step S104 with reference to figure 6, Fig. 6.Each character is extracted, first can be determined the position of image Chinese version image after described process by texture analysis, then Character segmentation is carried out to extract this each character to text image.
As shown in Figure 6, step S104 can comprise step S601 and step S602.
In step s 601, the position of the text image after described process in image is obtained.Can with reference to shown in figure 7, Fig. 7 is the particular flow sheet of step S601, specifically, can comprise the following steps:
Step S701, identifies the Edge texture in every row pixel by rim detection.Described Edge texture, refers to that in image, region jumpy occurs gray scale, can identify, that is, identify the region that grey scale change exceeds described predetermined threshold value variation range by arranging a predetermined threshold value variation range.
Step S702, does histogram to the Edge texture of every row pixel, and according to described histogrammic analysis being determined to the recognition threshold of edge primitive.Described edge primitive can be the pixel of gray-scale value in predetermined threshold range.The recognition threshold of described edge primitive, can carry out for utilizing adaptive thresholding algorithm the dynamic threshold calculated.
Step S703, edge primitive number in often going according to the recognition threshold statistics of described edge primitive, and record starting position and the end position of often row edge primitive.
Step S704, identifies the non-blank-white row in image after described process.Can according to the grey level histogram of image after described process, gray-scale value extreme difference (difference of gray-scale value maxima and minima) is identified as blank line lower than the row of predetermined threshold, and all the other are identified as non-blank-white row.Such as, by gray-scale value extreme difference lower than in histogram between maximum gradation value and minimum gradation value amplitude (extreme difference) 5% row be identified as blank line.In follow-up process using the blank line that identifies as blank background, do not do subsequent treatment, in subsequent treatment only using non-blank-white row as processing target.Wherein, predetermined threshold can be the variable obtained after basis is trained multiple samples pictures, such as, be directed to the license picture after current known training, predetermined threshold can be set in grey level histogram 5% of amplitude between maximum gradation value and minimum gradation value, for the picture recognition of other type, this variable can be configured according to the result of training other types image.
Step S705, judges whether current non-blank-white row meets pre-conditioned, if met, then carries out the detection of next non-blank-white row.Wherein, described pre-conditioned, a large amount of character sample can be sent in BP neural network and carry out training study, determine according to the result obtained after BP neural metwork training, such as, judge whether the number of the edge primitive in often going reaches predetermined number.
Step S706, when be consecutively detected the non-blank-white row exceeding predetermined number meet described pre-conditioned time, according to the starting position of the edge primitive of each non-blank-white row and the position of end position determination text image.
For the above-mentioned step S701-S706 determining the position of the text image after described process in image, execution sequence is not limited to above-mentioned one, other order can also be adopted to perform, such as, first can identify the non-blank-white row in image after described process, then the non-blank-white row identified be carried out to other step identification, judgement etc.
In step S602, Character segmentation is carried out to described text image, extract the image of each character in described text image.
Carrying out segmentation to described text image can utilize sciagraphy to carry out row cutting and character segmentation to extract the image of each character in described text image.Row cutting is exactly by the character cutting of a line a line out, forms single file character text image.The direction horizontal projection that can follow, by identifying that the blank between literal line and row carries out row cutting.Character segmentation, is carrying out row cutting and after the single file character text image obtained, is being cut out by single character picture, obtain the single character picture of each character from each single file character text image exactly.
In step S105, Text region is carried out to the image of each character extracted.
BP neural network can be utilized to carry out Text region to described character, the image of each character is sent into the identification carrying out character in BP nerve network system.
Wherein, in BP neural network in advance to the method for training the image array of character sample that the training of character sample can be passed through, namely, first the image of character sample is normalized, obtain the image array of each character sample, then BP neural network (error back propagation) training study is carried out to the image array of described each character sample.
When carrying out the identification of each character picture, the image of described each character is sent into the identification carrying out character in BP neural network.
Present invention also provides a kind of device identifying character on printed matter, Fig. 8 is the structured flowchart of the device 800 of character on the identification printed matter according to the embodiment of the present application, this device 800 can comprise as shown in the figure: acquisition module 810, hierarchical processing module 820, layer merges module 830, extraction module 840, and identification module 850.
Acquisition module 810 may be used for taking to obtain the image that will identify to described printed matter.
Hierarchical processing module 820 may be used for copying to obtain at least two width duplicating images to described image, and carries out different image procossing respectively to obtain at least two width layered images to every width duplicating image.
Layer merges module 830 and may be used for the layered image obtained to carry out layer merging, to obtain processing rear image.
Extraction module 840 may be used for the image extracting each character after described process image.
Identification module 850 may be used for carrying out character recognition to the image of each character extracted.
According to an embodiment of the application, described acquisition module 810 can be further used for carrying out exposure when taking by predetermined condition and arrange.
According to an embodiment of the application, described hierarchical processing module 820 can comprise denoising module and contrast-enhancement module.
Denoising module may be used for carrying out the process of removal noise to obtain first layer image to the width in described duplicating image.
Contrast-enhancement module may be used for carrying out contrast enhancement processing to obtain the second layered image to another width in described duplicating image.
According to an embodiment of the application, described denoising module can comprise: noise identification module, denoising gray-scale value acquisition module, and noise removes module.
Noise identification module may be used for identifying the noise in described duplicating image.
Denoising gray-scale value acquisition module may be used for using the gray-scale value of each noise with its around the gray-scale value phase adduction of adjacent eight pixels average as the denoising gray-scale value of each noise.
Noise is removed module and be may be used for the gray-scale value of noise each in described duplicating image to replace with the denoising gray-scale value of this noise to obtain first layer image.
According to an embodiment of the application, described noise identification module can comprise: calculating sub module, judges submodule, and recognin module.
Calculating sub module may be used for using the gray-scale value of pixel each in described duplicating image with its about the gray-scale value phase adduction of two neighbor pixels average as the calculating gray-scale value of each pixel.
Judge that submodule may be used for judging that the absolute value of the gray-scale value of each pixel and the difference of its calculating gray-scale value is whether in predetermined threshold range.
The pixel that recognin module may be used for gray-scale value and the absolute difference calculating gray-scale value exceed predetermined threshold range is identified as noise.
According to an embodiment of the application, described contrast-enhancement module can comprise picture portion module and gray scale adjusting module.
Picture portion module may be used for described duplicating image to be divided at least two sub regions.
Gray scale adjusting module may be used for carrying out gray scale adjustment respectively to every sub regions, to obtain the second layered image.
According to an embodiment of the application, described merging module 830 can comprise: value module and gray-scale value replacement module.
Value module may be used for getting intermediate value to the gray-scale value of pixel corresponding in described layered image, obtains the gray-scale value intermediate value of each pixel.
Gray-scale value replacement module may be used for the gray-scale value intermediate value gray-scale value of each pixel being replaced with this pixel, obtains processing rear image.
According to an embodiment of the application, described extraction module 840 can comprise:
Position acquisition module, may be used for obtaining the position of the text image after described process in image;
Character segmentation module, may be used for carrying out Character segmentation to described text image, extracts the image of each character in described text image.
According to an embodiment of the application, described position acquisition module may further include: edge detection module, threshold value acquisition module, statistic record module, non-blank-white row identification module, condition judgment module, and position determination module.
Edge detection module may be used for the Edge texture identified by rim detection in every row pixel.Wherein, described Edge texture can be the region of gray-scale value generation acute variation.
Threshold value acquisition module may be used for doing histogram to the Edge texture of every row pixel, and according to the recognition threshold to described histogram analysis determination edge primitive.
Statistic record module may be used for the quantity of adding up often row coboundary primitive according to the recognition threshold of described edge primitive, and records starting position and the end position of often row edge primitive.
Non-blank-white row identification module may be used for identifying the non-blank-white row after described process in image.
Condition judgment module may be used for judging whether current non-blank-white row meets pre-conditioned, if met, then carries out the detection of next non-blank-white row.
Position determination module may be used for when be consecutively detected the non-blank-white row exceeding predetermined number meet described pre-conditioned time, according to the starting position of the edge primitive of each non-blank-white row and the position of end position determination text image.
According to an embodiment of the application, described identification module 850 can be further used for, and utilizes the image of BP neural network to described each character to carry out character recognition.
The function realized due to the device of the present embodiment is substantially corresponding to the embodiment of the method shown in earlier figures 1 to Fig. 7, therefore not detailed part in the description of the present embodiment, see the related description in previous embodiment, can not repeat at this.
In one typically configuration, computing equipment comprises one or more processor (CPU), input/output interface, network interface and internal memory.
Internal memory may comprise the volatile memory in computer-readable medium, and the forms such as random access memory (RAM) and/or Nonvolatile memory, as ROM (read-only memory) (ROM) or flash memory (flash RAM).Internal memory is the example of computer-readable medium.
Computer-readable medium comprises permanent and impermanency, removable and non-removable media can be stored to realize information by any method or technology.Information can be computer-readable instruction, data structure, the module of program or other data.The example of the storage medium of computing machine comprises, but be not limited to phase transition internal memory (PRAM), static RAM (SRAM), dynamic RAM (DRAM), the random access memory (RAM) of other types, ROM (read-only memory) (ROM), Electrically Erasable Read Only Memory (EEPROM), fast flash memory bank or other memory techniques, read-only optical disc ROM (read-only memory) (CD-ROM), digital versatile disc (DVD) or other optical memory, magnetic magnetic tape cassette, tape magnetic rigid disk stores or other magnetic storage apparatus or any other non-transmitting medium, can be used for storing the information can accessed by computing equipment.According to defining herein, computer-readable medium does not comprise non-temporary computer readable media (transitory media), as data-signal and the carrier wave of modulation.
Also it should be noted that, term " comprises ", " comprising " or its any other variant are intended to contain comprising of nonexcludability, thus make to comprise the process of a series of key element, method, commodity or equipment and not only comprise those key elements, but also comprise other key elements clearly do not listed, or also comprise by the intrinsic key element of this process, method, commodity or equipment.When not more restrictions, the key element limited by statement " comprising ... ", and be not precluded within process, method, commodity or the equipment comprising described key element and also there is other identical element.
It will be understood by those skilled in the art that the embodiment of the application can be provided as method, system or computer program.Therefore, the application can adopt the form of complete hardware embodiment, completely software implementation or the embodiment in conjunction with software and hardware aspect.And the application can adopt in one or more form wherein including the upper computer program implemented of computer-usable storage medium (including but not limited to magnetic disk memory, CD-ROM, optical memory etc.) of computer usable program code.
The foregoing is only the embodiment of the application, be not limited to the application.To those skilled in the art, the application can have various modifications and variations.Any amendment done within all spirit in the application and principle, equivalent replacement, improvement etc., within the right that all should be included in the application.
Claims (20)
1. identify a method for character on printed matter, it is characterized in that, comprising:
Take described printed matter to obtain the image that will identify;
Copy to obtain at least two width duplicating images to described image, and different image procossing is carried out respectively to obtain at least two width layered images to every width duplicating image;
The layered image obtained is carried out layer merging, to obtain processing rear image;
The image of each character is extracted image after described process; And
Character recognition is carried out to the image of each character extracted.
2. method according to claim 1, is characterized in that, takes to obtain the image that will identify, comprising described printed matter: carry out exposure when taking by predetermined condition and arrange.
3. method according to claim 1, is characterized in that, carries out different image procossing respectively to obtain at least two width layered images, comprising each width duplicating image described:
The process of removal noise is carried out to obtain first layer image to the width in described duplicating image; And
Contrast enhancement processing is carried out to obtain the second layered image to another width in described duplicating image.
4. method according to claim 3, is characterized in that, carries out the process of removal noise to obtain first layer image, comprising the width in described duplicating image:
Identify the noise in described duplicating image;
Using the gray-scale value of each noise with its around the gray-scale value phase adduction of adjacent eight pixels average as the denoising gray-scale value of each noise; And
The gray-scale value of noise each in described duplicating image is replaced with the denoising gray-scale value of this noise to obtain first layer image.
5. method according to claim 4, is characterized in that, identifies that the noise in described duplicating image comprises:
Using the gray-scale value of pixel each in described duplicating image with its about the gray-scale value phase adduction of two neighbor pixels average as the calculating gray-scale value of each pixel;
Judge that the absolute value of the gray-scale value of each pixel and the difference of its calculating gray-scale value is whether in predetermined threshold range; And
The pixel that gray-scale value and the absolute value of the difference calculating gray-scale value exceed predetermined threshold range is identified as noise.
6. method according to claim 3, is characterized in that, carries out contrast enhancement processing to obtain the second layered image, comprising another width in described duplicating image:
Described duplicating image is divided at least two sub regions; And
Respectively gray scale adjustment is carried out to every sub regions, to obtain the second layered image.
7. method according to claim 1, is characterized in that, is merged by described layered image, to obtain processing rear image, comprising:
Intermediate value is got to the gray-scale value of pixel corresponding in described layered image, obtains the gray-scale value intermediate value of each pixel; And
The gray-scale value of each pixel is replaced with the gray-scale value intermediate value of this pixel, to obtain processing rear image.
8. method according to claim 1, is characterized in that, extracts the image of each character in image after described process, comprising:
Obtain the position of the text image after described process in image; And
Character segmentation is carried out to described text image, extracts the image of each character in described text image.
9. method according to claim 8, is characterized in that, obtains the position of the text image after described process in image, comprising:
The Edge texture in every row pixel is identified by rim detection;
Histogram is done to the Edge texture of every row pixel, and according to described histogrammic analysis being determined to the recognition threshold of edge primitive;
According to the number of the edge primitive that the recognition threshold statistics of described edge primitive is often gone, and record starting position and the end position of often row edge primitive;
Identify the non-blank-white row in image after described process;
Judge whether current non-blank-white row meets pre-conditioned, if met, then carry out the detection of next non-blank-white row; And
When be consecutively detected the non-blank-white row exceeding predetermined number meet described pre-conditioned time, according to the starting position of the edge primitive of each non-blank-white row and the position of end position determination text image.
10. method according to claim 1, is characterized in that, carries out character recognition, comprising the image of each character extracted: utilize the image of BP neural network to described each character to carry out character recognition.
11. 1 kinds of devices identifying character on printed matter, is characterized in that, comprising:
Acquisition module, obtains the image that will identify for taking described printed matter;
Hierarchical processing module, obtains at least two width duplicating images for copying described image, and carries out different image procossing respectively to obtain at least two width layered images to every width duplicating image;
Layer merges module, carries out layer merging, to obtain processing rear image for the layered image that will obtain;
Extraction module, for extracting the image of each character in image after described process; And
Identification module, for carrying out character recognition to the image of each character extracted.
12. devices according to claim 11, is characterized in that, described acquisition module, are further used for carrying out exposure when taking by predetermined condition and arrange.
13. devices according to claim 11, is characterized in that, described hierarchical block comprises:
Denoising module, for carrying out the process of removal noise to obtain first layer image to the width in described duplicating image; And
Contrast-enhancement module, for carrying out contrast enhancement processing to obtain the second layered image to another width in described duplicating image.
14. devices according to claim 13, is characterized in that, described denoising module, comprising:
Noise identification module, for identifying the noise in described duplicating image;
Denoising gray-scale value acquisition module, for using the gray-scale value of each noise with its around the gray-scale value phase adduction of adjacent eight pixels average as the denoising gray-scale value of each noise; And
Noise removes module, for the gray-scale value of noise each in described duplicating image being replaced with the denoising gray-scale value of this noise to obtain first layer image.
15. devices according to claim 14, is characterized in that, described noise identification module comprises:
Calculating sub module, for using the gray-scale value of pixel each in described duplicating image with its about the gray-scale value phase adduction of two neighbor pixels average as the calculating gray-scale value of each pixel;
Judge submodule, for the absolute value of the difference of the gray-scale value and its calculating gray-scale value that judge each pixel whether in predetermined threshold range; And
Recognin module, is identified as noise for the pixel that gray-scale value and the absolute value of the difference calculating gray-scale value are exceeded predetermined threshold range.
16. devices according to claim 13, is characterized in that, described contrast-enhancement module comprises:
Picture portion module, for being divided at least two sub regions by described duplicating image; And
Gray scale adjusting module, for carrying out gray scale adjustment respectively to every sub regions, to obtain the second layered image.
17. devices according to claim 11, is characterized in that, described merging module comprises:
Value module, for getting intermediate value to the gray-scale value of pixel corresponding in described layered image, obtains the gray-scale value intermediate value of each pixel; And
Gray-scale value replacement module, for the gray-scale value of each pixel being replaced with the gray-scale value intermediate value of this pixel, obtains processing rear image.
18. devices according to claim 11, is characterized in that, described extraction module comprises:
Position acquisition module, for obtaining the position of the text image after described process in image; And
Character segmentation module, for carrying out Character segmentation to described text image, extracts the image of each character in described text image.
19. devices according to claim 18, is characterized in that, described position acquisition module comprises:
Edge detection module, for identifying the Edge texture in every row pixel by rim detection;
Threshold value acquisition module, for doing histogram to the Edge texture of every row pixel, and according to the recognition threshold to described histogram analysis determination edge primitive;
Statistic record module, for adding up the quantity of often row coboundary primitive according to the recognition threshold of described edge primitive, and records starting position and the end position of often row edge primitive;
Non-blank-white row identification module, for identifying the non-blank-white row after described process in image;
Condition judgment module, for judging whether current non-blank-white row meets pre-conditioned, if met, then carries out the detection of next non-blank-white row; And
Position determination module, for when be consecutively detected the non-blank-white row exceeding predetermined number meet described pre-conditioned time, according to the starting position of the edge primitive of each non-blank-white row and the position of end position determination text image.
20. devices according to claim 11, is characterized in that, described identification module is further used for, and utilize the image of BP neural network to described each character to carry out character recognition.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310331468.4A CN104346609B (en) | 2013-08-01 | 2013-08-01 | The method and device of character on a kind of identification printed matter |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310331468.4A CN104346609B (en) | 2013-08-01 | 2013-08-01 | The method and device of character on a kind of identification printed matter |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104346609A true CN104346609A (en) | 2015-02-11 |
| CN104346609B CN104346609B (en) | 2018-05-04 |
Family
ID=52502183
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310331468.4A Active CN104346609B (en) | 2013-08-01 | 2013-08-01 | The method and device of character on a kind of identification printed matter |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104346609B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104978578A (en) * | 2015-04-21 | 2015-10-14 | 深圳市前海点通数据有限公司 | Mobile phone photo taking text image quality evaluation method |
| CN105160304A (en) * | 2015-08-10 | 2015-12-16 | 中山大学 | Method and device for sign text identification based on machine vision |
| CN105787480A (en) * | 2016-02-26 | 2016-07-20 | 广东小天才科技有限公司 | Method and device for shooting test questions |
| CN107145734A (en) * | 2017-05-04 | 2017-09-08 | 深圳市联新移动医疗科技有限公司 | A kind of automatic acquisition of medical data and input method and its system |
| CN107545460A (en) * | 2017-07-25 | 2018-01-05 | 广州智选网络科技有限公司 | One kind digitlization color page promotion management and analysis method, storage device and mobile terminal |
| CN110135288A (en) * | 2019-04-28 | 2019-08-16 | 佛山科学技术学院 | A kind of quick checking method and device of electronics license |
| CN110929738A (en) * | 2019-11-19 | 2020-03-27 | 上海眼控科技股份有限公司 | Certificate card edge detection method, device, equipment and readable storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2006031163A (en) * | 2004-07-13 | 2006-02-02 | Ricoh Co Ltd | Character recognition result processor, character recognition result processing method, character recognition result processing program and recording medium with the same program stored |
| CN1756311A (en) * | 2004-09-29 | 2006-04-05 | 乐金电子（惠州）有限公司 | Image switching method and its apparatus |
| CN102289792A (en) * | 2011-05-03 | 2011-12-21 | 北京云加速信息技术有限公司 | Method and system for enhancing low-illumination video image |
| CN102663382A (en) * | 2012-04-25 | 2012-09-12 | 重庆邮电大学 | Video image character recognition method based on submesh characteristic adaptive weighting |
-
2013
- 2013-08-01 CN CN201310331468.4A patent/CN104346609B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2006031163A (en) * | 2004-07-13 | 2006-02-02 | Ricoh Co Ltd | Character recognition result processor, character recognition result processing method, character recognition result processing program and recording medium with the same program stored |
| CN1756311A (en) * | 2004-09-29 | 2006-04-05 | 乐金电子（惠州）有限公司 | Image switching method and its apparatus |
| CN102289792A (en) * | 2011-05-03 | 2011-12-21 | 北京云加速信息技术有限公司 | Method and system for enhancing low-illumination video image |
| CN102663382A (en) * | 2012-04-25 | 2012-09-12 | 重庆邮电大学 | Video image character recognition method based on submesh characteristic adaptive weighting |
Non-Patent Citations (5)
| Title |
|---|
| SOU6: ""Photoshop照片模糊变清晰大全"", 《HTTP://WWW.31IAN.COM/EDU/2012/04-03/24407.HTML?&FROM=ANDROIDQQ》 * |
| ZHOULPWEN: ""快速提高照片清晰度"", 《HTTP://JINGYAN.BAIDU.COM/ARTIC1E/FEC4BCE20AE348F2608D8B64.HTML》 * |
| 小照: ""PS几种处理模糊照片变清晰的方法"", 《HTTP://WWW.31IAN.COM/EDU/2012/07-21/32980.HTML》 * |
| 张惠: ""同源视频检索与商标货号识别"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 益彩足球: ""利用PS把不清晰的照片改清晰"", 《HTTP://JINGYAN.BAIDU.COM/ARTIC1E/F3AD7D0FDC433A09C3345B0B.HTML》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104978578A (en) * | 2015-04-21 | 2015-10-14 | 深圳市前海点通数据有限公司 | Mobile phone photo taking text image quality evaluation method |
| CN104978578B (en) * | 2015-04-21 | 2018-07-27 | 深圳市点通数据有限公司 | Mobile phone photograph text image method for evaluating quality |
| CN105160304A (en) * | 2015-08-10 | 2015-12-16 | 中山大学 | Method and device for sign text identification based on machine vision |
| CN105787480A (en) * | 2016-02-26 | 2016-07-20 | 广东小天才科技有限公司 | Method and device for shooting test questions |
| CN105787480B (en) * | 2016-02-26 | 2020-01-03 | 广东小天才科技有限公司 | Method and device for shooting test questions |
| CN107145734A (en) * | 2017-05-04 | 2017-09-08 | 深圳市联新移动医疗科技有限公司 | A kind of automatic acquisition of medical data and input method and its system |
| CN107145734B (en) * | 2017-05-04 | 2020-08-28 | 深圳市联新移动医疗科技有限公司 | Automatic medical data acquisition and entry method and system |
| CN107545460A (en) * | 2017-07-25 | 2018-01-05 | 广州智选网络科技有限公司 | One kind digitlization color page promotion management and analysis method, storage device and mobile terminal |
| CN110135288A (en) * | 2019-04-28 | 2019-08-16 | 佛山科学技术学院 | A kind of quick checking method and device of electronics license |
| CN110929738A (en) * | 2019-11-19 | 2020-03-27 | 上海眼控科技股份有限公司 | Certificate card edge detection method, device, equipment and readable storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104346609B (en) | 2018-05-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104346609A (en) | Method and device for recognizing characters on printed products | |
| CN110414507B (en) | License plate recognition method and device, computer equipment and storage medium | |
| CN109670500B (en) | Text region acquisition method and device, storage medium and terminal equipment | |
| CN112686812B (en) | Bank card inclination correction detection method and device, readable storage medium and terminal | |
| US9053540B2 (en) | Stereo matching by census transform and support weight cost aggregation | |
| CN108108734B (en) | License plate recognition method and device | |
| CN110020692B (en) | Handwriting separation and positioning method based on print template | |
| CN108197644A (en) | A kind of image-recognizing method and device | |
| CN109447117B (en) | Double-layer license plate recognition method and device, computer equipment and storage medium | |
| CN110738238B (en) | Classification positioning method and device for certificate information | |
| CN114387591A (en) | License plate recognition method, system, equipment and storage medium | |
| CN113191358B (en) | Metal part surface text detection method and system | |
| CN111814673A (en) | Method, device and equipment for correcting text detection bounding box and storage medium | |
| CN112991536A (en) | Automatic extraction and vectorization method for geographic surface elements of thematic map | |
| Siddiqui et al. | Block-based feature-level multi-focus image fusion | |
| CN110321887B (en) | Document image processing method, document image processing apparatus, and storage medium | |
| CN111881938B (en) | Image authenticity identification method based on blind forensics technology | |
| CN111797832B (en) | Automatic generation method and system for image region of interest and image processing method | |
| Shobha Rani et al. | Restoration of deteriorated text sections in ancient document images using atri-level semi-adaptive thresholding technique | |
| CN112488106A (en) | Fuzzy, inclined and watermark-carrying identity card copy element extraction method | |
| CN112686247A (en) | Identification card number detection method and device, readable storage medium and terminal | |
| Chang et al. | An efficient thresholding algorithm for degraded document images based on intelligent block detection | |
| CN112052859B (en) | License plate accurate positioning method and device in free scene | |
| CN115797327A (en) | Defect detection method and device, terminal device and storage medium | |
| CN106886777B (en) | Character boundary determining method and device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20191203 Address after: P.O. Box 31119, grand exhibition hall, hibiscus street, 802 West Bay Road, Grand Cayman, Cayman Islands Patentee after: Innovative advanced technology Co., Ltd Address before: A four-storey 847 mailbox in Grand Cayman Capital Building, British Cayman Islands Patentee before: Alibaba Group Holding Co., Ltd. |