CN110060265A - A method of divide from painting and calligraphy cultural relic images and extracts seal - Google Patents
A method of divide from painting and calligraphy cultural relic images and extracts seal Download PDFInfo
- Publication number
- CN110060265A CN110060265A CN201910401692.3A CN201910401692A CN110060265A CN 110060265 A CN110060265 A CN 110060265A CN 201910401692 A CN201910401692 A CN 201910401692A CN 110060265 A CN110060265 A CN 110060265A
- Authority
- CN
- China
- Prior art keywords
- seal
- image
- painting
- calligraphy
- generates
- 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.)
- Pending
Links
- 238000010422 painting Methods 0.000 title claims abstract description 45
- 238000000034 method Methods 0.000 title claims abstract description 31
- 239000000284 extract Substances 0.000 title claims abstract description 9
- 230000011218 segmentation Effects 0.000 claims abstract description 21
- 239000000203 mixture Substances 0.000 claims abstract description 4
- 238000010189 synthetic method Methods 0.000 claims abstract description 4
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 230000001133 acceleration Effects 0.000 abstract description 3
- 239000002131 composite material Substances 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 239000004615 ingredient Substances 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 230000001902 propagating effect Effects 0.000 description 1
- 238000004904 shortening Methods 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/50—Image enhancement or restoration using two or more images, e.g. averaging or subtraction
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30176—Document
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of from dividing in painting and calligraphy cultural relic images and extract the method for seal the following steps are included: step 1. by being scanned to the seal image in seal dictionary, constructs standard seal image database；Step 2. generates the prospect mark of Pixel-level for standard seal image database seal image by the method for image procossing；Step 3. generates seal mask images and is handled；Step 4. generates the seal image for having painting and calligraphy background by image synthetic method；Step 5. marks seal prospect and the seal image with painting and calligraphy background that is obtained by image composition method is as training sample, training semantic segmentation model；Step 6. is split and extracts to the seal in painting and calligraphy cultural relic images using trained semantic segmentation model.Influence of the painting and calligraphy pieces content to seal itself can be farthest excluded through the invention, the seal image of clean noiseless factor is obtained, identifies for seal, and the accuracy that seal identification not only can be improved can be restrained with acceleration model, shorten the training time.
Description
Technical field
The present invention relates to field of image processing more particularly to a kind of sides divided from painting and calligraphy cultural relic images and extract seal
Method.
Background technique
Seal in painting and calligraphy pieces conveyed the information of painting and calligraphy author, be the important side that spectators understand painting and calligraphy background information
Formula.But since seal is often using seal character word and many authors joined artistic wound when carving seal
The ingredient of work, has made lettering and has largely changed.Therefore, the General Visitors of relevant knowledge are not difficult to pick out capping
It is which word actually in seal.
So if can be shown by the seal in automatic identification painting and calligraphy pieces the information of seal owner for
General Visitors appreciate painting and calligraphy pieces, and the expansion information for understanding painting and calligraphy has huge help, for carrying forward and propagating national tradition
Culture has huge progradation.However, the seal covered on painting and calligraphy pieces is often overlapped with the content of works,
Since painting and calligraphy pieces content is affected to seal, directly the seal on painting and calligraphy is identified, accuracy rate is often very low, this
Huge challenge is constituted to the automatic identification technology of seal.Therefore, the present invention proposes that one kind is divided from painting and calligraphy cultural relic images
And the method for extracting seal, influence of the painting and calligraphy pieces content to seal itself can be farthest excluded, is obtained clean without dry
The seal image of factor is disturbed, is identified for seal, the accuracy that seal identification not only can be improved can be restrained with acceleration model
Shorten the training time.
Summary of the invention
Purpose to realize the present invention, is achieved using following technical scheme:
A method of divide from painting and calligraphy cultural relic images and extract seal, comprising the following steps:
Step 1. constructs standard seal image database by being scanned to the seal image in seal dictionary；
Step 2. generates the prospect of Pixel-level for standard seal image database seal image by the method for image procossing
Mark；
Step 3. generates seal mask images and is handled；
Step 4. generates the seal image for having painting and calligraphy background by image synthetic method；
Step 5. seal prospect is marked and the seal image with painting and calligraphy background that is obtained by image composition method as
Training sample, training semantic segmentation network model；
Step 6. is split and mentions to the seal in painting and calligraphy cultural relic images using trained semantic segmentation network model
It takes.
The method, in which: standard seal image database includes that the seal image of rgb format and seal text are said
It is bright.
The method, wherein step 2 include:
2.1 the seal image of rgb format is transformed to gray level image using following formula:
Igray=IR·0.299+IG·0.587+IB·0.114
Wherein, IgrayFor the gray level image after conversion, IR, IG, IBThe respectively red, green, blue channel of RGB image；
2.2 are calculated as follows the average value of gray level image as image threshold T:
Wherein, Igray(r, c) is the r row of grayscale image, the pixel value at c column, R, C be respectively grayscale image number of lines of pixels and
Columns；
2.3 according to threshold value T to gray level image IgrayIn each pixel proceed as follows, obtain prospect mark, assignment
Prospect is corresponded to for 1 pixel, and is assigned a value of 0 pixel and corresponds to background:
The method, wherein step 3 include:
3.1 mark I by duplication prospectgroundtruthObtain display foreground mask Imask；
3.2 pairs of display foreground masks do random noise and are superimposed to obtain Inoise-mask；；
Black white binarization is added noisy seal image foreground mask I by 3.3noise-maskBe converted to red seal figure
Picture:
The method, wherein step 4 include:
4.1 collect the image of painting and calligraphy historical relic works, and interception does not include the image of seal largely at random from painting and calligraphy image
Region, as background image；
4.2 have the seal image of painting and calligraphy background by following formula synthesis:
Icomb(r, c)=λ (r, c) Ired-mask(r,c)+(1-λ(r,c))·Ibackground(r,c)
Wherein, IbackgroundFor the image-region as painting and calligraphy background extracted at random from painting and calligraphy image, λ (r, c) is
The fusion coefficients when pixel of r row c column in the composite image, the coefficient are the random number greater than 0 less than 1.
The method, in which: the loss function that the training semantic segmentation network model in step 5 uses is as follows:
Wherein, piFor the true mark of pixel i in image, foreground pixel 1, background pixel 0,For prediction pixel i
Belong to the probability of prospect.
The method, wherein step 6 includes: that the seal image with painting and calligraphy background is inputted semantic segmentation model, language
Adopted parted pattern carries out semantic segmentation to input picture, and all pixels for being predicted to be prospect are extracted and are predicted as the figure of background
As being then dropped, seal image is obtained.
Detailed description of the invention
Fig. 1 is the flow diagram for dividing and extracting seal from painting and calligraphy cultural relic images.
Specific embodiment
The specific embodiment of the invention is further described with reference to the accompanying drawings and examples.
As shown in Figure 1, dividing from painting and calligraphy cultural relic images and the method for extracting seal includes:
The first step constructs standard seal image database by being scanned to the seal image in seal dictionary.Standard
It include the seal image and seal explanatory note of rgb format in seal image database；
Second step, by the method for image procossing, before generating Pixel-level for standard seal image database seal image
Scape mark.Include:
The seal image of rgb format is transformed to gray level image using following methods by 2.1:
Igray=IR·0.299+IG·0.587+IB·0.114
Wherein, IgrayFor the gray level image after conversion, IR, IG, IBThe respectively red, green, blue channel of RGB image.
2.2 are calculated as follows the average value of gray level image as image threshold:
Wherein, Igray(r, c) is the r row of grayscale image, the pixel value at c column, R, C be respectively grayscale image number of lines of pixels and
Columns.
2.3 according to threshold value T to gray level image IgrayIn each pixel proceed as follows, obtain prospect mark, assignment
Prospect is corresponded to for 1 pixel, and is assigned a value of 0 pixel and corresponds to background:
Third step generates seal mask images and carries out noise superposition and data extending.Include:
3.1 mark I by duplication prospectgroundtruthObtain display foreground mask Imask；
3.2 pairs of display foreground masks do random noise superposition: the process is the partial pixel randomly selected in image, will be selected
The foreground pixel in pixel taken is set to background pixel and the background pixel of selection is set to foreground pixel, obtains plus noisy
Foreground mask Inoise-mask.The effect of the more close practical seal image of the image obtained by way of the plus noise.
Black white binarization is added noisy seal image foreground mask I by 3.3noise-maskBe converted to red seal figure
Picture:
Additionally, it is preferred that can further include following steps: I is marked to red seal image and prospectgroundtruth
It does identical rotation process and obtains the seal data that rotation is expanded；By carrying out brightness fluctuation to the pixel in red seal image
Obtain the image of brightness fluctuation expansion；In addition it is also possible to increase other data extending methods, for example, to red seal image into
Row Fuzzy processing, it is hereby achieved that large number of red seal image sample.
4th step generates the seal image for having painting and calligraphy background by image synthetic method.
4.1 collect the image of painting and calligraphy historical relic works, and interception does not include the image of seal largely at random from painting and calligraphy image
Region, as background image；
4.2 are synthesized by the following the seal image with painting and calligraphy background:
Icomb(r, c)=λ (r, c) Ired-mask(r,c)+(1-λ(r,c))·Ibackground(r,c)
Wherein, IbackgroundFor the image-region as painting and calligraphy background extracted at random from painting and calligraphy image, λ (r, c) is
The fusion coefficients when pixel of r row c column in the composite image, the coefficient are the random number greater than 0 less than 1.
5th step, the band that the seal prospect obtained by image processing method is marked and obtained by image composition method
The seal image of painting and calligraphy background is as training sample, training semantic segmentation network model.Here semantic segmentation model only needs
Differentiation belongs to the foreground pixel of seal and is not belonging to the background pixel of seal.Segmentation network model can be used but not limited to use
Full convolution semantic segmentation network or with coding-decoding structure SegNet semantic segmentation network.Training semantic segmentation network mould
The loss function that type uses is as follows:
Wherein, piFor the true mark of pixel i in image, foreground pixel 1, background pixel 0,For prediction pixel i
Belong to the probability of prospect.Training is iterated to semantic segmentation model based on back-propagation algorithm.
6th step, after completing training, semantic segmentation model can be used for the segmentation of the seal in painting and calligraphy historical relic and extract.
In this course, the seal image with painting and calligraphy background is entered semantic segmentation model, semantic segmentation model schemes input
As carrying out semantic segmentation.The pixel that all pixels for being predicted to be prospect are extracted and are predicted as background is then dropped, and is printed
Chapter image.
Extraction is split to seal image using method proposed by the present invention, can farthest exclude painting and calligraphy pieces
Influence of the content to seal itself obtains the seal form of clean noiseless factor.It is mentioned using based on method proposed by the present invention
The seal taken carries out seal identification, not only can farthest exclude influence of the contextual factor to seal and improve identification accurately
Rate can also enable disaggregated model preferably learn the validity feature of seal, acceleration model convergence, when shortening the training of model
Between.
Claims (3)
1. a kind of method divided from painting and calligraphy cultural relic images and extract seal, it is characterised in that the following steps are included:
Step 1. constructs standard seal image database by being scanned to the seal image in seal dictionary；
Step 2. generates the prospect mark of Pixel-level for standard seal image database seal image by the method for image procossing
Note；
Step 3. generates seal mask images and is handled；
Step 4. generates the seal image for having painting and calligraphy background by image synthetic method；
Step 5. marks seal prospect and the seal image with painting and calligraphy background that is obtained by image composition method is as training
Sample, training semantic segmentation network model；
Step 6. is split and extracts to the seal in painting and calligraphy cultural relic images using trained semantic segmentation network model.
2. according to the method described in claim 1, it is characterized by: standard seal image database includes the seal of rgb format
Image and seal explanatory note.
3. according to the method described in claim 1, it is characterized in that step 2 includes:
The seal image of rgb format is transformed to gray level image using following formula by 2.1:
Igray=IR·0.299+IG·0.587+IB·0.114
Wherein, IgrayFor the gray level image after conversion, IR, IG, IBThe respectively red, green, blue channel of RGB image.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910401692.3A CN110060265A (en) | 2019-05-15 | 2019-05-15 | A method of divide from painting and calligraphy cultural relic images and extracts seal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910401692.3A CN110060265A (en) | 2019-05-15 | 2019-05-15 | A method of divide from painting and calligraphy cultural relic images and extracts seal |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110060265A true CN110060265A (en) | 2019-07-26 |
Family
ID=67323157
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910401692.3A Pending CN110060265A (en) | 2019-05-15 | 2019-05-15 | A method of divide from painting and calligraphy cultural relic images and extracts seal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110060265A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112766275A (en) * | 2021-04-08 | 2021-05-07 | 金蝶软件（中国）有限公司 | Seal character recognition method and device, computer equipment and storage medium |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104954741A (en) * | 2015-05-29 | 2015-09-30 | 东方浩联（北京）智能科技有限公司 | Tramcar on-load and no-load state detecting method and system based on deep-level self-learning network |
| CN107145846A (en) * | 2017-04-26 | 2017-09-08 | 贵州电网有限责任公司输电运行检修分公司 | A kind of insulator recognition methods based on deep learning |
| CN107256550A (en) * | 2017-06-06 | 2017-10-17 | 电子科技大学 | A kind of retinal image segmentation method based on efficient CNN CRF networks |
| CN107274345A (en) * | 2017-06-07 | 2017-10-20 | 众安信息技术服务有限公司 | A kind of Chinese printable character image combining method and device |
| CN107330889A (en) * | 2017-07-11 | 2017-11-07 | 北京工业大学 | A kind of traditional Chinese medical science tongue color coating colour automatic analysis method based on convolutional neural networks |
| CN107679502A (en) * | 2017-10-12 | 2018-02-09 | 南京行者易智能交通科技有限公司 | A kind of Population size estimation method based on the segmentation of deep learning image, semantic |
| CN107679465A (en) * | 2017-09-20 | 2018-02-09 | 上海交通大学 | A kind of pedestrian's weight identification data generation and extending method based on generation network |
| CN108010034A (en) * | 2016-11-02 | 2018-05-08 | 广州图普网络科技有限公司 | Commodity image dividing method and device |
| CN108932735A (en) * | 2018-07-10 | 2018-12-04 | 广州众聚智能科技有限公司 | A method of generating deep learning sample |
| CN109284758A (en) * | 2018-09-29 | 2019-01-29 | 武汉工程大学 | A kind of invoice seal removing method, device and computer storage medium |
| CN109325532A (en) * | 2018-09-18 | 2019-02-12 | 成都网阔信息技术股份有限公司 | The image processing method of EDS extended data set under a kind of small sample |
| CN109614983A (en) * | 2018-10-26 | 2019-04-12 | 阿里巴巴集团控股有限公司 | The generation method of training data, apparatus and system |
-
2019
- 2019-05-15 CN CN201910401692.3A patent/CN110060265A/en active Pending
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104954741A (en) * | 2015-05-29 | 2015-09-30 | 东方浩联（北京）智能科技有限公司 | Tramcar on-load and no-load state detecting method and system based on deep-level self-learning network |
| CN108010034A (en) * | 2016-11-02 | 2018-05-08 | 广州图普网络科技有限公司 | Commodity image dividing method and device |
| CN107145846A (en) * | 2017-04-26 | 2017-09-08 | 贵州电网有限责任公司输电运行检修分公司 | A kind of insulator recognition methods based on deep learning |
| CN107256550A (en) * | 2017-06-06 | 2017-10-17 | 电子科技大学 | A kind of retinal image segmentation method based on efficient CNN CRF networks |
| CN107274345A (en) * | 2017-06-07 | 2017-10-20 | 众安信息技术服务有限公司 | A kind of Chinese printable character image combining method and device |
| CN107330889A (en) * | 2017-07-11 | 2017-11-07 | 北京工业大学 | A kind of traditional Chinese medical science tongue color coating colour automatic analysis method based on convolutional neural networks |
| CN107679465A (en) * | 2017-09-20 | 2018-02-09 | 上海交通大学 | A kind of pedestrian's weight identification data generation and extending method based on generation network |
| CN107679502A (en) * | 2017-10-12 | 2018-02-09 | 南京行者易智能交通科技有限公司 | A kind of Population size estimation method based on the segmentation of deep learning image, semantic |
| CN108932735A (en) * | 2018-07-10 | 2018-12-04 | 广州众聚智能科技有限公司 | A method of generating deep learning sample |
| CN109325532A (en) * | 2018-09-18 | 2019-02-12 | 成都网阔信息技术股份有限公司 | The image processing method of EDS extended data set under a kind of small sample |
| CN109284758A (en) * | 2018-09-29 | 2019-01-29 | 武汉工程大学 | A kind of invoice seal removing method, device and computer storage medium |
| CN109614983A (en) * | 2018-10-26 | 2019-04-12 | 阿里巴巴集团控股有限公司 | The generation method of training data, apparatus and system |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112766275A (en) * | 2021-04-08 | 2021-05-07 | 金蝶软件（中国）有限公司 | Seal character recognition method and device, computer equipment and storage medium |
| CN112766275B (en) * | 2021-04-08 | 2021-09-10 | 金蝶软件（中国）有限公司 | Seal character recognition method and device, computer equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Buslaev et al. | Fully convolutional network for automatic road extraction from satellite imagery | |
| CN108416377B (en) | Information extraction method and device in histogram | |
| CN100550038C (en) | Image content recognizing method and recognition system | |
| CN110880000B (en) | Picture character positioning method and device, computer equipment and storage medium | |
| CN108364278B (en) | Rock core fracture extraction method and system | |
| CN106022363A (en) | Method for recognizing Chinese characters in natural scene | |
| CN110516554A (en) | A kind of more scene multi-font Chinese text detection recognition methods | |
| CN105719247A (en) | Characteristic learning-based single image defogging method | |
| CN109214285A (en) | Detection method is fallen down based on depth convolutional neural networks and shot and long term memory network | |
| CN109544468B (en) | Image data amplification method | |
| CN112907598B (en) | Method for detecting falsification of document and certificate images based on attention CNN | |
| CN110503103A (en) | A kind of character cutting method in line of text based on full convolutional neural networks | |
| CN110472628A (en) | A kind of improvement Faster R-CNN network detection floating material method based on video features | |
| CN112686276A (en) | Flame detection method based on improved RetinaNet network | |
| CN110598698A (en) | Natural scene text detection method and system based on adaptive regional suggestion network | |
| CN110880176A (en) | Semi-supervised industrial image defect segmentation method based on countermeasure generation network | |
| WO2021034841A1 (en) | Apparatus and methods for converting lineless tables into lined tables using generative adversarial networks | |
| CN107730508A (en) | Color documents images multichannel binary processing method | |
| CN110399760A (en) | A kind of batch two dimensional code localization method, device, electronic equipment and storage medium | |
| CN112819837A (en) | Semantic segmentation method based on multi-source heterogeneous remote sensing image | |
| CN108711160A (en) | A kind of Target Segmentation method based on HSI enhancement models | |
| CN111462090A (en) | Multi-scale image target detection method | |
| CN115205626A (en) | Data enhancement method applied to field of coating defect detection | |
| CN110060265A (en) | A method of divide from painting and calligraphy cultural relic images and extracts seal | |
| CN113436198A (en) | Remote sensing image semantic segmentation method for collaborative image super-resolution reconstruction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190726 |