CN101295359B - Image processing method and image processing apparatus - Google Patents

Image processing method and image processing apparatus Download PDF

Info

Publication number
CN101295359B
CN101295359B CN2008100058810A CN200810005881A CN101295359B CN 101295359 B CN101295359 B CN 101295359B CN 2008100058810 A CN2008100058810 A CN 2008100058810A CN 200810005881 A CN200810005881 A CN 200810005881A CN 101295359 B CN101295359 B CN 101295359B
Authority
CN
China
Prior art keywords
mentioned
information
data
background
color
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.)
Expired - Fee Related
Application number
CN2008100058810A
Other languages
Chinese (zh)
Other versions
CN101295359A (en
Inventor
关峰伸
浅野英辅
永吉洋登
永崎健
新庄广
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hitachi Omron Financial System Co Ltd
Original Assignee
Hitachi Omron Financial System Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Hitachi Omron Financial System Co Ltd filed Critical Hitachi Omron Financial System Co Ltd
Publication of CN101295359A publication Critical patent/CN101295359A/en
Application granted granted Critical
Publication of CN101295359B publication Critical patent/CN101295359B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/20Analysis of motion
    • G06T7/254Analysis of motion involving subtraction of images
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/90Determination of colour characteristics

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Multimedia (AREA)
  • Character Input (AREA)
  • Image Analysis (AREA)
  • Image Processing (AREA)
  • Facsimile Image Signal Circuits (AREA)

Abstract

The invention provides an image processing program and an image processing device, recorded words, prints, marks and the like are extracted in a high accuracy from document images comprising color bias or fuzzy color. The image processing method of the invention is composed of: a background removing and data generating part for removing background parts from the color images or bright images, generating, and generating the background removing data which represents the parts other than the background; a profile color converting and data generating part for, in the parts other than the background in the color images or bright images, generating the data which converts the profile color of the parts other than the background into pixel color at inner side of the profile of the parts other than the background.

Description

Image processing apparatus and image processing method
Technical field
The present invention relates to a kind of OCR (OCR:OpticalCharacter Reader), scanner, digital camera etc. of utilizing and for example take documents such as account ticket, from the file and picture that generates, extract image processing method and the image processing apparatus of charging to special objects such as literal, trace, mark.
Background technology
In financial institution or autonomous body, use scanner devices such as OCR, realize the high efficiency of document process business such as account ticket.The major function of OCR is generation, the extraction of file and picture Chinese words, the literal identification of file and picture.Kind as the file and picture that generates has bianry image, luminance picture, coloured image.
Use the processing of bianry image, because data volume is little, so the processing time tails off.But in binary Images Processing, preprinted ruling, bit line, prompt text, shade and the literal of charging to hand-written or printing afterwards of preprinting (Preprint) of being called has under the very big overlapping situation in the account ticket, is difficult to distinguish them.Therefore, there is the situation of the part shortcoming of situation that produces noise among the extraction result of literal or the literal that extracts, the problem that has literal identification to make mistakes.
The processing of using luminance picture is that the shading image of black and white is handled.Because in luminance picture is handled, utilize to preprint and differently with the brightness value of charging to literal distinguish them, therefore preprinting and charging under the overlapping situation of literal, distinguish them and become easy than binary Images Processing.But under the close situation of the brightness value of preprinting and charge to literal, their discrimination precision step-down.
In the processing that utilizes coloured image, owing to can distinguish according to preprinting with the different of the color of charging to literal, therefore distinguishing their specific luminance Flame Image Process becomes easy.In Color Image Processing, the color of preprinting by removal extracts charges to literal, trace, mark etc.
In this method, the method of removing the removal color of appointment before the input of account ticket as [patent documentation 3] is arranged, and as [patent documentation 1] or [patent documentation 2], extract specific shape part in the account ticket of input as the ruling, remove the method that extracts color same color partly with this.
[patent documentation 1] spy opens 2003-196592
[patent documentation 2] spy opens 2005-258683
[patent documentation 3] spy opens 2006-134355
[patent documentation 4] spy opens 2004-336106
[patent documentation 5] spy opens 2005-18810
In above-mentioned Color Image Processing, exist owing to produce colo(u)r bias in the image by OCR, scanner, digital camera generation, can not correctly extract special objects such as charging to literal or trace, preprint and leave a part, or the problems such as part shortcoming of special object.
So-called colo(u)r bias is meant the color component of 3 primary colors that sense, become red component the R value, become the G value of green component and become the offset of at least one value in the B value of blue component.As the main cause that colo(u)r bias produces, enumerate the allocation position, transporting velocity of aberration, the sensor of camera lens etc.Particularly in the scanner of the two-dimensional CCD that utilizes desktop scanners or digital camera etc., produce the more colo(u)r bias of bringing because of aberration.
Because colo(u)r bias, in the outline portion of the special object of preprinting or charge to literal etc., produce the false colour different with the original color of special object.For example, have in the profile of black literal, produce the situation of red and blue false colour, or on the profile of the ruling of blueness, produce the situation etc. of ruddy false colour.Therefore, according to the information of color difference charge to literal and the Color Image Processing preprinted etc. in produce mistake.
To this, [patent documentation 41 is attempted removing aberration, [patent documentation 5] of camera lens and is attempted removing the colo(u)r bias that is produced by the allocation position of sensor.
Yet even count the method for side and revisal departure, it is difficult removing colo(u)r bias from image fully.In addition, the problem of the ruling time that more high-precision colo(u)r bias revisal will be expensive has also occurred.
In addition, in above-mentioned Color Image Processing or luminance picture processing, produce in image under the fuzzy situation of color, special objects such as charging to literal or trace can not be correctly extracted in existence, leave a part to preprint, or the problem of the part of literal shortcoming.
So-called color is fuzzy, is meant ruling or charges to the color of outline portion of literal fuzzy, produces light color.Make the dispersion of colouring informations such as the red component of preprinting or charge to literal, blue component, green component, lightness, chroma, form and aspect, brightness become big owing to color is fuzzy, therefore difference is charged to literal and is preprinted the difficulty that becomes.
Summary of the invention
The present invention finishes in view of these problems, provides a kind of from contain the fuzzy file and picture of colo(u)r bias or color, extracts image processing method and the image processing apparatus of charging to special objects such as literal, trace, mark accurately.
For achieving the above object, the present invention is from utilizing scanner or digital camera to read the coloured image or luminance picture of documents such as account ticket, extraction is charged in the image processing method of special objects such as literal, trace, mark, possesses following feature, have: remove background from coloured image or luminance picture, the background of the part beyond the generation display background is removed the background of data and is removed the generation processing; The profile color conversion data that are created on beyond the above-mentioned background in coloured image or the luminance picture in the part, the colouring information of the profile of part beyond the background is converted to the data of the colouring information of the image of the profile inboard of part beyond background generate to be handled; Extract processing with the special object that extracts the special object part.
In addition, above-mentioned profile color conversion data generate to handle and are characterised in that, for the concerned pixel in coloured image or the brightness original image, with reference near pixel, generate the low-light level color expansion brightness data that colouring informations such as the red component of concerned pixel, blue component, green component, lightness, chroma, form and aspect, brightness is converted to the color of pixel information that brightness value is minimum in the nearby pixel and concerned pixel as in its vicinity a plurality of pixels.
Above-mentioned special object juggling is characterised in that, carry out that ruling extracts, the special object candidate extracts, the colouring information of the colouring information of ruling and special object infer differentiation with special object.
According to the present invention, even colo(u)r bias or fuzzy coloured image or the luminance picture of color are arranged, also can distinguish special objects such as preprinting, charge to literal, trace, mark accurately, for example can only extract accurately and charge to literal.Be not limited only to charge to literal, also can extract the special objects in file and picture such as trace or mark accurately.
Description of drawings
Fig. 1 is the figure that the expression special object extracts the structure of handling.
Fig. 2 is the figure of presentation video treating apparatus.
Fig. 3 is the example of coloured image.
Fig. 4 is that background is removed data.
Fig. 5 is the differentiation result of special object.
Fig. 6 is that the expression background is removed the figure that data generate the example of handling.
Fig. 7 is the figure that represents special object juggling in the past.
Fig. 8 is that ruling extracts the result.
Fig. 9 is that ruling is removed the result.
Figure 10 is that the special object candidate extracts the result.
Figure 11 is an example of charging to the colo(u)r bias of literal.
Figure 12 is the example of the colo(u)r bias of ruling.
Figure 13 is the figure that expression profile color conversion data generate the example of handling.
Figure 14 is that the profile color conversion data of the image of expression Figure 11 generate the figure that handles.
Figure 15 is that the profile color conversion data of the image of expression Figure 12 generate the figure that handles.
Figure 16 is the figure of expression special object juggling.
Figure 17 is the figure that represents only to utilize the special object juggling of inferring of ruling color.
Figure 18 is the figure that represents only to utilize the special object juggling of inferring of special object color.
Figure 19 is the figure that expression utilizes the special object juggling of cluster.
Figure 20 is the figure that the special object of expression interpolation colo(u)r bias revisal extracts the structure of handling procedure.
Figure 21 is that expression possesses the figure that the special object that extracts the object color specifying function extracts the structure of handling procedure.
Figure 22 is that expression comprises the figure that specifies the subordinate special object juggling program that extracts the object color.
Figure 23 is the figure that expression utilization comprises the special object juggling of specifying the cluster that extracts the object color.
Figure 24 is the figure of the fuzzy example of expression color.
Figure 25 is the situation about handling of figure expression generates to(for) the profile color conversion data that the fuzzy image of color is arranged.
Figure 26 is the figure that the profile color conversion data of expression luminance picture input generate the example of handling.
Figure 27 is the demonstration example of coloured image.
Figure 28 is the differentiation result's of special object a demonstration example.
(symbol description)
101 images are obtained module
102 backgrounds are removed data generation module
103 profile color conversion data generation module
104 special object discrimination modules
105 control modules
Embodiment
Below, for using image processing method of the present invention and image processing apparatus, utilize accompanying drawing to describe in detail.
Embodiment 1
Fig. 2 is the figure of an embodiment of expression image processing apparatus of the present invention.
This is the image processing apparatus 208 that input medias 207 such as communicator 201, image capturing device 202, display device 203, external memory 204, storer 205, CPU (Central Processing Unit) 206, keyboard or mouse is utilized order wire connections such as pci bus.
The program of the structure that possesses special object extraction processing shown in Figure 1 is contained in the memory storage of external memory 204 or storer 205 etc., utilizes CPU205 to carry out.
Be input to the coloured image or the luminance picture of the document among the CPU, can also can be stored in the external memory 204 from image capturing devices such as scanner, OCR 202 or communicator 201 inputs.
The result that special object extract to be handled has situation in display device of outputing to 203, outputs to outside situation or be used to the situation etc. of other programs in image is obtained treating apparatus 208 via communicator 201.As the example of other programs, the program of carrying out literal identification is arranged.
Figure 27 is with from 201 inputs of image capturing device 202 or communicator or be stored in the example that shows in the display window 2701 of coloured image on display device 203 external memory 204.In addition, Figure 28 extracts the example that shows in the display window 2702 of result on display device 203 of handling with special object.
Fig. 1 is that the figure that special object of the present invention extracts the structure of handling procedure is used in expression.Special object extract handling procedure by image obtain module 101, background is removed data generation module 102, profile color conversion data generation module 103, special object discrimination module 104 and control module 105 and is constituted.
Image is obtained module 101 and is utilized scanner or OCR etc. to obtain the coloured image of image conversions such as paper document or the image of luminance picture are obtained processing.
Background is removed data generation module 102 and is carried out generation background in coloured image from be input to CPU206 or the luminance picture and remove the background of data and remove data and generate and handle.
For example, obtaining under the situation of the coloured image of preprinting and charge to literal 304 that image pattern 3 contains ruling 301, bit line 302 and shade 303 like that, background is removed the data that data generation module 102 generates demonstration ruling, bit line shown in Figure 4 and charges to word segment.
It is the background parts of removing in the image that background is removed data generation processing, extracts the processing of preprinting partly and charge to word segment of ruling and bit line.For realizing that this processing has several different methods, takes method shown in Figure 6.
At first, generate in the processing 601, from the coloured image of representing by 3 primary colors (R value, G value, B value) of RGB, generate the luminance picture of representing by brightness in the brightness value data.Then, cut apart (block generation) at piece and handle in 602, luminance picture is divided into a plurality of.At last, in binary conversion treatment 603, each piece is created on pixel in the piece that brightness value is low is made as the two-value data that black, pixel that brightness value is high are made as white.The two-value data of Sheng Chenging so as shown in Figure 4, is that black picture element represents that the background of the part beyond the background removes data.
The profile color conversion data that profile color conversion data generation module 103 generates profile color conversion data 1303 generate processing, these profile color conversion data 1303 are that input color image 604 and background are removed data 605, are that the color of the inside part of profile obtains with ruling, bit line and the color conversion of charging to the profile of literal.In addition, coloured image 604 can be a luminance picture.
Special object discrimination module 104, carry out for be input to background among the CPU206 remove data 605, with reference to profile color conversion data 1303, generate the determination processing of the special object of the expression data of charging to word segment shown in Figure 5, the output special object is differentiated result 706.
Here, utilize Fig. 7 that in the past special object juggling is described.In special object juggling in the past, the input background is removed data, with reference to coloured image, and the differentiation result of output special object.
Fig. 7 represents special object juggling in the past.At first, extract to handle 701 at ruling ' in, extract the ruling part.In this was handled, the part of removing the long rectilinearity longways of the black picture element ground connection in the data by the extraction background extracted the ruling part.Fig. 8 consequently.
Then, remove at ruling and to handle 702 ' in, generate and remove the ruling of having removed the ruling part the data from background and remove data.Fig. 9 consequently.
Then, extract to handle 703 in the special object candidate ' in, remove the data from ruling, utilize the size of rectangle or the information of position, extract to become and charge to the word segment candidate as the candidate of charging to word segment of special object.Figure 10 consequently.
Then, handle 704 inferring of ruling color and special object color ' in, by reference coloured image 604, infer as the ruling part colouring information of the colouring information of ruling part and charge to literal candidate part colouring information as the colouring information of charging to literal candidate part.
Then, the juggling 705 of special object ' in, utilize ruling part colouring information and charge to literal candidate part colouring information, differentiate background and remove whether each pixel of the black picture element part in the data is the pixel of charging to literal.This processing is to remove in each locations of pixels of the black picture element part in the data in background, and the colouring information of differentiating coloured image belongs to ruling part colouring information, still belongs to the processing of the colouring information of charging to literal candidate part.
Specifically, each pixel of removing the black picture element part in the data 605 in background is carried out following processing.Remove certain black picture element position (Xa in the data 605 in background, Xb) in the processing, judgement coloured image 604 (Xa, Xb) colouring information in handle the ruling part colouring information of 704 ' output by inferring of ruling color and special object color and charge in the literal candidate part colouring information which approaching.And, if (Xa, colouring information Xb) is near ruling part colouring information, then judge (Xa, position Xb) is the ruling part, if (Xa, Xb) colouring information judges then that near charging to literal candidate part colouring information (Xa, position Xb) is to charge to word segment.
As the colouring information of this coloured image 604, can utilize R value, G value, the B value of RGB 3 primary colors, also can be the colouring information that they have been changed, for example form and aspect, chroma, the lightness in brightness value or HSV space.A value in them can be only utilized in addition, also a plurality of values can be utilized.In addition, in method of discrimination, can use the multiple differentiation algorithm that utilizes teacher's data.For example, utilize neural network, linear identification device, mahalanobis distance (MahalanobisDistance) etc.
Then,, carry out the determination processing of special object by reference coloured image 604, output special object differentiation result 706 ', the special object juggling finishes.
Yet, under the situation of in the past special object juggling, if in the coloured image 604 of input colo(u)r bias is arranged, because the colouring information in the coloured image 604 also produces deviation, so, therefore have and utilize colouring information can not distinguish the problem of preprinting and charge to literal based on the possibility that produces deviation being arranged near the ruling part colouring information and which the judgement itself of colouring information charged in the literal candidate part colouring information.Therefore, according to the special object result of determination of utilizing previous methods to obtain, can not obtain the differentiation result of the present application sometimes, for example such output of Fig. 5.
Here, Figure 11 is the example of image (charging to literal) that colo(u)r bias is arranged.In Figure 11, be to produce blue false colour and red false colour on the profile of charging to literal of black originally.
In addition, Figure 12 also is the example of image (ruling) that colo(u)r bias is arranged.In Figure 12, be to have produced ruddy false colour on the profile of blue ruling originally.Consideration utilizes colouring information only to extract the situation of charging to literal from the image of charging to the literal and the ruling of blueness that comprises black.
In charging to literal and ruling, do not have under the situation of colo(u)r bias, can utilize in the past special object juggling only to extract and charge to literal.But, as Figure 11 and Figure 12, having in the image of colo(u)r bias, owing in the profile of the profile of charging to literal and ruling, all there is a red component, the situation of the situation that the outline portion that therefore ruling arranged produces as noise or the part shortcoming of literal.Producing under the situation of colo(u)r bias like this, the problem that can not utilize the colouring information difference to preprint and charge to literal is being arranged.
For Figure 11,12 such images, in using image processing apparatus of the present invention, carry out profile color conversion data and generate and handle, carry out the special object juggling with reference to profile color conversion data.
Above-mentioned profile color conversion data generation module 103 is carried out profile color conversion data and is generated processing.Specifically, be created on beyond the background in the coloured image 604 in the part, with the color conversion of the profile of part beyond the background data for the color of pixel in the inboard of the profile of part beyond background.That is to say, generate the data of colouring information that the ruling in the coloured image, bit line and the colouring information of charging to the profile of literal are converted to the inside part of this profile.
Figure 13 is the example that profile color conversion data generate the concrete treatment scheme of handling.
Generate in the processing in profile color conversion data, via storer 205, input color image 604 and background are removed data 605 from communicator 201, image capturing device 202 or external memory 204.
And nearby brightness value generates in processing 1301 and the low-light level color expansion process 1302, selects (selecteed pixel is called concerned pixel) pixel in the zone beyond the background in coloured image one by one, changes the colouring information of this concerned pixel.All pixels that these two processing repeat in the zone to the background in coloured image are processed.
Nearby brightness value generate to be handled in 1302, generates the brightness value of pixel near the field of concerned pixel (in the example at Figure 14, as with the concerned pixel being 9 pixels in the field of 3 * 3 scope at center) respectively.With the pixel divided into beyond the concerned pixel in the field is near pixel.Above-mentioned field is not limited only to 3 * 3, for example also can be 2 * 2 or 4 * 4.In addition, concerned pixel is not limited only to the center in the field, the field can be set at any position that makes concerned pixel be positioned at the field.
Then, in low-light level color expansion process 1302, with the colouring information (for example R value, G value and B value) of concerned pixel be converted to concerned pixel and near the minimum color of pixel information of brightness value in the pixel.Like this, the colouring information that R value, G value, B value are produced the profile portion of deviation is converted to the colouring information of the inboard of profile portion, becomes the processing that simulation ground is converted to false colour original colouring information.
More particularly, calculate in the field concerned pixel and near the brightness value of pixel, extract pixel with minimum brightness value, the colouring information of concerned pixel is converted to the color of pixel information with minimum brightness value.If the brightness value of concerned pixel is minimum brightness value, the colouring information of concerned pixel is in statu quo kept.Like this, the ruling in coloured image 604, bit line and charge in the part of literal generate as the profile color conversion data 1303 of having changed the data of colouring information.
Utilize profile color conversion data generate to handle, for example charging under the situation of literal, as shown in figure 14, will be converted to the low black of brightness in the profile inboard in the high redness of the brightness in the profile portion of charging to literal shown in Figure 11 and blue false colour.
In addition, utilize profile color conversion data generate to handle, for example under the situation of ruling, as shown in figure 15, will be converted to the low blueness of brightness in this profile inboard in the high ruddy false colour of the brightness in the profile portion of the ruling shown in Figure 12.
Figure 16 is the figure of the concrete treatment scheme of special object juggling in the present embodiment 1.
At first, import the ruling extraction processing 701 that background is removed data, extracted the ruling part.
Then, generate and remove the ruling of removing ruling part the data from background and remove the ruling of data and remove and handle 702.
Then, carry out from ruling remove the data, utilize the size of rectangle or position information, extract the special object candidate of charging to the word segment candidate that becomes as the candidate of charging to word segment of special object and extract and handle 703.
Then, in special object juggling of the present invention, in the juggling 1602 of inferring processing 1601 and special object of ruling color and special object color, with reference to the rgb value of profile color conversion data 1303.
Remove in the zone of profile color conversion data 1303 of black-pixel region of data being equivalent to background color, because having the pixel of the false colour that produces because of colo(u)r bias tails off, therefore special object color and ruling color to infer precision better, as a result of, also improve the precision of the differentiation of ruling and special object.
Like this, in using image processing apparatus 208 of the present invention, ruling color and special object color infer 1601 and the juggling 1602 of special object in be made as black owing to can will charge to literal portion, ruling portion is made as blueness and handles, and therefore can correctly differentiate and charge to word segment.
More than, according to image processing apparatus 208, reference contains the profile color conversion data of the rgb value after the profile color conversion processing, therefore can extract the literal of charging to that becomes special object accurately from the coloured image that contains colo(u)r bias.In addition, will extract the character recognition device of result, and can obtain more high-precision recognition result as the literal of charging to of the output of this image processing apparatus as input.And, literal is charged in extraction is utilized as an example, but under the situation that extracts trace or mark, can extract accurately too.
Below, other embodiments of the present invention are described.
Embodiment 2
As shown in figure 17, also can in special object judegment part 104, adopt the special object extraction processing that only utilizes inferring of ruling color and carry out the differentiation of special object.
Processing shown in Figure 17 is to handle in 1701 inferring of ruling color, with reference to profile color conversion data, only infers the colouring information of ruling.Then, handle in 1702,, remove data 605 from background and remove ruling color part, differentiate the word segment of charging to that becomes special object by utilizing the colouring information of ruling removing of ruling color part.
Embodiment 3
As shown in figure 18, also can in special object judegment part 104, adopt the special object extraction processing that only utilizes inferring of special object color and carry out the special object juggling.
Processing shown in Figure 180 is to handle in 1801 inferring of special object color, with reference to profile color conversion data 1303, only infers the colouring information of special object candidate.Then, handle in 1702, utilize the colouring information of special object, remove data 605 from background and extract the word segment of charging to that becomes special object in the extraction of special object color part.
Embodiment 4
As shown in figure 19, also can adopt in special object judegment part 104, the special object that utilizes cluster to carry out the special object juggling extracts to be handled.
In the processing shown in Figure 19, do not utilize the result that ruling extracts and only utilize beyond the background colouring information partly to differentiate.At first in clustering processing 1901, the profile color conversion data 1303 of part beyond the background are carried out cluster.In cluster, can utilize R value, G value, the B value of RGB 3 primary colors, also can be the colouring information that they have been changed, for example form and aspect, chroma, the lightness in brightness value or HSV space.A value in them can be only utilized in addition, also a plurality of values can be utilized.In the method for cluster, k-means method or zone broadening method or discriminant analysis methods are arranged.
Then, handle in 1902, from utilize a plurality of classes that cluster obtains, select the class of special object in the selection of the class of special object.The method of selecting has several different methods, for example selects to have the methods such as class of the high value of brightness value.
And, in the extraction 1903 of certain object classes color part,, extract the literal of charging to that becomes special object by extracting the pixel of the colouring information of class the black picture element part of removing data from background with above-mentioned selection.
Embodiment 5
Also can adopt at special object shown in Figure 1 and extract in the structure of handling procedure, the special object that adds colo(u)r bias revisal module 2001 again extracts to be handled.
It is as shown in figure 20 structure that this special object extracts handling procedure, carries out except following illustrated processing and as shown in Figure 1 the identical processing of embodiment.
Colo(u)r bias revisal module 2001 is carried out the colo(u)r bias revisal and is handled.The colo(u)r bias revisal is handled by change and is utilized file and picture to obtain R value, G value, the B value of handling obtained coloured image 604, or expansion dwindles etc., generates the colo(u)r bias revisal data of data that conduct has alleviated the deviation of color.
And, with respect in structure shown in Figure 1, being input to background removes the data that data generate to be handled, profile color conversion data generate in handling and utilizes coloured image 604, among the embodiment in Figure 20, being input to background, to remove the data that data generate to be handled, profile color conversion data generate in handling are colo(u)r bias revisal data.Like this, even in the many images of the departure of colo(u)r bias, also can extract special objects such as charging to literal accurately.
Embodiment 6
Also can adopt at special object shown in Figure 1 and extract in the structure of handling procedure, add the special object extraction processing that designated color is obtained module 2101 again.
This special object extracts and handles is as shown in figure 21 structure, carries out except following illustrated processing and as shown in Figure 1 the identical processing of embodiment.
Obtain in the module at designated color, carry out designated color and obtain processing.Designated color is obtained and is obtained in the processing as the special object that extracts and the color of appointment is promptly specified and extracted object colouring information 2203.Specify to extract the object colouring information about this, user's specified message or from information of input medias such as keyboard or mouse input or the like in program is in advance arranged.And this colouring information can utilize R value, G value, the B value of RGB, also can be the colouring information that they have been changed, for example form and aspect, chroma, the lightness in brightness value or HSV space.A value in them can be only utilized in addition, also a plurality of values can be utilized.In addition, can be the value that shows a color, also can be the scope of the value of Show Color.
And the special object juggling becomes the processing that will specify extraction object colouring information 2203 to be included in the input as Figure 22 or Figure 23.
Figure 22 utilize to specify extracts 1601 the result of inferring of object colouring information 2203 and ruling color and special object color in the special object juggling of Figure 16, carry out the differentiation 2201 of special object.
Figure 23 in the special object juggling of the cluster 1901 of utilizing Figure 19, utilize to specify extract object colouring information 2203 carry out special object class determine 2301.
Embodiment discussed above, not only for the problem of colo(u)r bias, the problem fuzzy for color also is effective.Figure 24 is that the outline portion of charging to literal becomes the fuzzy example of light color.For the image of Figure 24, generate processing if carry out profile color conversion data, just generate profile color conversion data 1303 shown in Figure 25.In profile color conversion data 1303, the part of fuzzy light color is converted into dark color in the coloured image 604 of input.Like this, for there being the fuzzy image of color also can extract special object accurately.
In addition, embodiment discussed above is not only under the situation of input color document, for also being effective having imported under the situation that produces the fuzzy luminance picture of color.Under the situation of having imported luminance picture, in the embodiment in figure 1, by profile color conversion data generation processing being made as processing shown in Figure 26 and can tackling.
Input luminance picture 2604 and background are removed data 605 among Figure 26, in the part in luminance picture 2604 except background, each pixel are carried out one by one the processing of the low-light level color expansion process 2601 of luminance picture.And the part beyond the background just at ruling, bit line with charge in the part of literal, generates the profile color conversion data 1303 as the data that brightness value in the luminance picture 2604 has been changed.
In the low-light level color expansion process 2601 of luminance picture, the brightness value of the pixel that brightness value is minimum in concerned pixel and near the pixel is converted to the brightness value of concerned pixel.

Claims (14)

1. image processing apparatus is characterized in that possessing:
Remove background from the image information of input, the background that generates expression background zone is in addition removed the unit of data;
Generate the unit of color conversion data, these color conversion data are in the zone beyond the background of image information of input, the color conversion data that the color of pixel information translation of the profile relevant with special object is obtained for the color of pixel information in the inboard of above-mentioned profile;
The storage above-mentioned background is removed the unit of data and above-mentioned color conversion data; With
Remove the candidate that data are selected special object from above-mentioned background, with reference to above-mentioned color conversion data, the special object judgement unit of output special object part.
2. image processing apparatus according to claim 1 is characterized in that:
Above-mentioned image information is the color image information or the monochrome information of image.
3. image processing apparatus according to claim 1 is characterized in that:
In the unit of above-mentioned generation color conversion data,
The background in the zone beyond the expression above-mentioned background is removed the field of selecting to comprise a plurality of pixels the data,
Generate the brightness value of the pixel in the above-mentioned field,
Generation is converted to the minimum color of pixel information of brightness value in the pixel in above-mentioned field with the colouring information of the concerned pixel in the above-mentioned field and the color conversion data that obtain.
4. image processing apparatus according to claim 1 is characterized in that:
Above-mentioned image information is the information in the external memory, the information of utilizing image capturing device to obtain of being stored in, or from the information of communicator input any.
5. image processing apparatus according to claim 1 is characterized in that:
The unit of above-mentioned generation color conversion data,
Set the concerned pixel on the above-mentioned image information, with reference near pixel as the pixel near the field of concerned pixel,
Generation is converted to the minimum color of pixel information of brightness value in the nearby pixel and concerned pixel with the colouring information of above-mentioned concerned pixel and the low-light level color expansion brightness data that obtains.
6. image processing apparatus according to claim 1 is characterized in that:
Above-mentioned colouring information is in red component, blue component, green component, lightness, chroma, form and aspect, the brightness at least more than any one.
7. image processing apparatus according to claim 1 is characterized in that also possessing:
Remove the unit that extracts ruling information the data from above-mentioned background; With
Remove the unit of removing the ruling information that extracts the data from above-mentioned background,
Wherein, above-mentioned special object judgement unit is removed the candidate of selecting special object the data from the background of having removed above-mentioned ruling information, with reference to above-mentioned color conversion data, and output special object part.
8. image processing method, utilize possess storage unit, treating apparatus that image is obtained unit, display unit, this image processing method possesses:
Remove background from the image information that is obtained the unit input by above-mentioned image, the background that generates expression background zone is in addition removed the step of data;
Generate the step of color conversion data, these color conversion data are in the zone beyond the background of image information of input, the color conversion data that the color of pixel information translation of the profile relevant with special object is obtained for the color of pixel information in the inboard of above-mentioned profile;
Above-mentioned background is removed data and the step of color conversion data storage in storage unit; With
Background from be stored in said memory cells is removed in the data candidate of selecting special object, with reference to above-mentioned color conversion data, special object is partly outputed to special object discriminating step in the above-mentioned display unit.
9. image processing method according to claim 8 is characterized in that:
Above-mentioned image information is the color image information or the monochrome information of image.
10. image processing method according to claim 8 is characterized in that:
In the step of above-mentioned generation color conversion data,
The background of the image information beyond the expression above-mentioned background is removed the field of selecting to comprise a plurality of pixels the data,
The brightness value of the pixel in the generation field,
Generation is converted to minimum brightness value in the pixel in the field with the brightness value of the concerned pixel in the field and the color conversion data that obtain.
11. image processing method according to claim 8 is characterized in that:
Above-mentioned image information is the information in the external memory of being stored in, utilize information that image capturing device obtains or any from the information of communicator input.
12. image processing method according to claim 8 is characterized in that:
In the step of above-mentioned generation color conversion data,
For the concerned pixel in the field, with reference near pixel as the pixel beyond the concerned pixel in the field,
Generation is converted to the minimum color of pixel information of brightness value in the nearby pixel and concerned pixel with the colouring information of above-mentioned concerned pixel and the low-light level color expansion brightness data that obtains.
13. image processing method according to claim 8 is characterized in that:
Above-mentioned colouring information is in red component, blue component, green component, lightness, chroma, form and aspect, the brightness at least more than any one.
14. image processing method according to claim 8 is characterized in that, also possesses:
Remove the step that extracts ruling information the data from above-mentioned background; With
Remove the step of removing the ruling information that extracts the data from above-mentioned background,
Wherein, in above-mentioned special object discriminating step, remove the candidate of selecting special object the data from the background of removing above-mentioned ruling information, with reference to above-mentioned color conversion data, output special object part.
CN2008100058810A 2007-04-25 2008-02-15 Image processing method and image processing apparatus Expired - Fee Related CN101295359B (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2007114988A JP4857173B2 (en) 2007-04-25 2007-04-25 Image processing apparatus, image processing method, and image processing program
JP2007-114988 2007-04-25

Publications (2)

Publication Number Publication Date
CN101295359A CN101295359A (en) 2008-10-29
CN101295359B true CN101295359B (en) 2010-09-29

Family

ID=40048876

Family Applications (1)

Application Number Title Priority Date Filing Date
CN2008100058810A Expired - Fee Related CN101295359B (en) 2007-04-25 2008-02-15 Image processing method and image processing apparatus

Country Status (4)

Country Link
JP (1) JP4857173B2 (en)
KR (1) KR101461233B1 (en)
CN (1) CN101295359B (en)
TW (1) TWI350997B (en)

Families Citing this family (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP5127739B2 (en) * 2009-02-06 2013-01-23 キヤノン株式会社 Image processing method, image processing apparatus, and program
JP5337563B2 (en) * 2009-04-08 2013-11-06 日立コンピュータ機器株式会社 Form recognition method and apparatus
CN101916327B (en) * 2010-07-09 2011-11-09 北京商纳科技有限公司 Method and system for generating wrong answer list
JP5867045B2 (en) * 2011-12-12 2016-02-24 富士ゼロックス株式会社 Image processing apparatus and program
RU2534005C2 (en) * 2013-02-01 2014-11-27 Корпорация "САМСУНГ ЭЛЕКТРОНИКС Ко., Лтд." Method and system for converting screenshot into metafile
JP6250793B2 (en) * 2014-04-16 2017-12-20 グローリー株式会社 Method and apparatus for removing character background from color image
JP6779688B2 (en) * 2016-07-25 2020-11-04 キヤノン株式会社 Image processing equipment, image processing method, computer program
CN106228157B (en) * 2016-07-26 2018-03-27 江苏鸿信系统集成有限公司 Coloured image word paragraph segmentation and recognition methods based on image recognition technology
CN106599818B (en) * 2016-12-07 2020-10-27 广州视源电子科技股份有限公司 Method and device for generating handwriting format file based on picture
US10602026B2 (en) * 2017-06-20 2020-03-24 Fuji Xerox Co., Ltd. Image processing apparatus, image processing method, and non-transitory computer readable medium
JP6958096B2 (en) * 2017-08-10 2021-11-02 富士フイルムビジネスイノベーション株式会社 Information processing equipment and programs
JP7059799B2 (en) 2018-05-23 2022-04-26 富士フイルムビジネスイノベーション株式会社 Information processing equipment and programs
CN110895696A (en) * 2019-11-05 2020-03-20 泰康保险集团股份有限公司 Image information extraction method and device
CN113083804A (en) * 2021-04-25 2021-07-09 中国铁建重工集团股份有限公司 Laser intelligent derusting method and system and readable medium

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1674035A (en) * 2004-03-22 2005-09-28 佳能株式会社 Image-processing device and image processing method
US20060023943A1 (en) * 2004-07-30 2006-02-02 Canon Kabushiki Kaisha Image processing method and image processing apparatus
US20060228044A1 (en) * 2005-04-12 2006-10-12 Newsoft Technology Corporation Method for automatically cropping image objects
CN1848929A (en) * 2005-04-15 2006-10-18 三星电子株式会社 AV (audio/video) system and method for forming bitmap font outline of the AV system

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS58142675A (en) * 1982-02-18 1983-08-24 Sanyo Electric Co Ltd Color picture processing system
JP3048158B2 (en) * 1988-10-04 2000-06-05 キヤノン株式会社 Color image processing equipment
JP2746692B2 (en) * 1989-10-09 1998-05-06 富士通株式会社 Color image data processing device
JPH0414960A (en) * 1990-05-09 1992-01-20 Fujitsu Ltd Color picture reader
JPH06266816A (en) * 1993-03-12 1994-09-22 Fujitsu Ltd Method and device for processing color picture
JP3923293B2 (en) * 2000-11-22 2007-05-30 シャープ株式会社 Image processing method, image processing apparatus, and image forming apparatus
JP4141310B2 (en) * 2003-04-16 2008-08-27 株式会社リコー Image processing apparatus, image processing method, and program executed by computer
JP4127691B2 (en) * 2004-10-04 2008-07-30 株式会社東芝 Character recognition apparatus and method

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1674035A (en) * 2004-03-22 2005-09-28 佳能株式会社 Image-processing device and image processing method
US20060023943A1 (en) * 2004-07-30 2006-02-02 Canon Kabushiki Kaisha Image processing method and image processing apparatus
US20060228044A1 (en) * 2005-04-12 2006-10-12 Newsoft Technology Corporation Method for automatically cropping image objects
CN1848929A (en) * 2005-04-15 2006-10-18 三星电子株式会社 AV (audio/video) system and method for forming bitmap font outline of the AV system

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
石洗凡.基于彩色图像处理技术的车辆牌照自动读取系统.计算机工程与科学.2005,27(9),47-50. *

Also Published As

Publication number Publication date
JP4857173B2 (en) 2012-01-18
JP2008269509A (en) 2008-11-06
KR101461233B1 (en) 2014-11-12
TW200842734A (en) 2008-11-01
CN101295359A (en) 2008-10-29
KR20080095743A (en) 2008-10-29
TWI350997B (en) 2011-10-21

Similar Documents

Publication Publication Date Title
CN101295359B (en) Image processing method and image processing apparatus
CN107633239B (en) Bill classification and bill field extraction method based on deep learning and OCR
US9508001B2 (en) Note recognition and management using color classification
Kanagarathinam et al. Text detection and recognition in raw image dataset of seven segment digital energy meter display
JP5701181B2 (en) Image processing apparatus, image processing method, and computer program
JP5830338B2 (en) Form recognition method and form recognition apparatus
CN107659799B (en) Image pickup apparatus, image processing method, and storage medium
CN104978565B (en) A kind of pictograph extracting method of universality
CN104408449A (en) Intelligent mobile terminal scene character processing method
CN104008401A (en) Method and device for image character recognition
CN106203415A (en) A kind of bank based on Digital Image Processing card number automatic identification equipment
CN106446885A (en) Paper-based Braille recognition method and system
CN103530625A (en) Optical character recognition method based on digital image processing
JPH11306325A (en) Method and device for object detection
Patil et al. Multi font and size optical character recognition using template matching
JP2014071556A (en) Image processor, image processing method, and program
CN110619331A (en) Color distance-based color image field positioning method
CN113065407B (en) Financial bill seal erasing method based on attention mechanism and generation countermeasure network
CN101620668A (en) Image recognition processing system and method thereof for capturing near-monochromatic character features
JP3955467B2 (en) Image processing program and image processing apparatus
CN114581922A (en) Chinese handwritten signature recognition method and system based on deep learning
KR101288754B1 (en) Apparatus for recognizing color channel code
Karthick et al. Consumer service number recognition using template matching algorithm for improvements in ocr based energy consumption billing
Ghanem et al. Impact of color space and color resolution on vehicle recognition models
CN109961020A (en) A kind of invoice seal minimizing technology

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20100929

Termination date: 20180215