CN106407969A - Robust complex background video text positioning and extracting method - Google Patents
Robust complex background video text positioning and extracting method Download PDFInfo
- Publication number
- CN106407969A CN106407969A CN201610778073.2A CN201610778073A CN106407969A CN 106407969 A CN106407969 A CN 106407969A CN 201610778073 A CN201610778073 A CN 201610778073A CN 106407969 A CN106407969 A CN 106407969A
- Authority
- CN
- China
- Prior art keywords
- text
- angle point
- area
- positioning
- region
- 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
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
- G06V10/225—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 based on a marking or identifier characterising the area
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/752—Contour matching
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
- Character Input (AREA)
Abstract
The invention discloses a robust complex background video text positioning and extracting method, and mainly solves the problems that the present video text positioning and extracting method is not sufficiently robust under the complex background. In the text positioning phase, angular points are selected to act as the basic characteristics of characters, and then positioning of text rows is realized by the method of combination of coarse positioning and fine positioning. Candidate text areas are obtained in coarse positioning by utilizing the four characteristics of Area, Saturation, Ratio and Position of a binary text area distribution graph. Finally, the candidate text areas are segmented into the text rows by using the horizontal projection of the angular points and the method of angular point density fusion so that accurate positioning of the text rows can be realized and pseudo text rows can be eliminated. The text extracting phase is completed through polarity judgment based on binary images and an improved local OTSU method so that the problems of optimal threshold selection and stroke detail loss under the complex background can be effectively solved. The robust complex background video text positioning and extracting method has high recall rate in video text positioning and extracting of various programs.
Description
Technical field
The invention belongs to field of video image processing is and in particular to a kind of healthy and strong complex background Video Text Location and taking out
Take method.
Background technology
With the fast development of modern science and technology, a lot of information in life are all to be transmitted by multimedia form.Wherein,
Word in video is one of most useful information type, and these texts provide much valuable information, and such as program is situated between
Continue, scene location, especially bulletin, the title of speaker, contest scores, date and time, Real Estate Trend, media event and regard
Frequency content etc..Text identification has had a lot of real world applications, such as visual classification, document analysis, the video inspection based on video content
Rope, help blind person, automatic marking, Car license recognition etc..So extracting to the text message of video, to the deep layer understanding video
Semantic information is significant.
There has been now a lot of video text positioning and the algorithm extracting both at home and abroad, can be largely classified into based on connected domain
, based on texture, based on edge and based on study method.Method locating speed ratio wherein based on connected domain is very fast,
But easily by disturbing that picture contrast changes；Method positioning based on texture and edge is more stable, but has the time complicated
Spend high shortcoming；And the method positioning quality based on study depends entirely on the training of sample.
Content of the invention
The present invention is directed in prior art the unhealthy and strong problem of Video Text Location under complex background it is proposed that a kind of video
The positioning of middle complex background text and extraction robust method.
The purpose of the present invention is achieved through the following technical solutions：A kind of healthy and strong complex background Video Text Location
And abstracting method, comprise the following steps：
Pretreatment：Input video frame, frame of video is converted into gray level image, carries out Corner Detection to frame of video, obtains angle
Point two-value scattergram；
Text coarse positioning：First with angle point two-value scattergram, region merging technique is carried out by sliding window, form two-value literary composition
One's respective area scattergram, then filters out the angle point in non-textual region using four attributes of two-value text regional distribution chart, real
Existing text filed coarse positioning；
Text fine positioning：By the method for angle point floor projection and angle point density fusion, realize line of text fine positioning；
Text extracts：The polarity that line of text image after positioning is carried out based on bianry image judges, then using improvement
Local OTSU algorithm calculate threshold value in each piece of region, and optimal threshold is adjusted to according to polarity judged result, finally complete
Become the binaryzation of line of text.
Further, through region, fusion obtains described two-value text regional distribution chart, specifically selects a n*n's
Rectangle frame, 5≤n≤15, centered on angle point, angle point color is set to the color in rectangle frame region, has traveled through all angle points,
Finally give two-value text regional distribution chart.
Further, described two-value text regional distribution chart has multiple connected regions.
Further, the attribute of described two-value text regional distribution chart includes：Area, Saturation, Ratio and
Position.
Further, the described angle point filtering out in non-textual region refers to by four attribute character Area,
Saturation, Ratio and Position are progressively filtered, specifically：
First, Area attribute filter refer to each of current two-value text regional distribution chart connected region Area by
It is ranked up according to size, filter out relatively small region Area；Then, Saturation filters and refers to calculate each even
The ratio of the area A (Rect) of area A (Area) and boundary rectangle frame Rect of logical region Area, Saturation=A
(Area)/A (Rect), (Saturation ∈ (0,1)), filter out the less connected region of Saturation；Then, Ratio
Filter and refer to calculate the depth-width ratio of each connected region boundary rectangle frame, filter out Ratio and be more than 1：2.5 connected region；?
Afterwards, Position filter method specifically calculates the positional information of connected region, filters out the connected region of part in frame of video 2/3
Domain.
Further, the method for described angle point floor projection and angle point density fusion refers to：
First, the angle point rectangular histogram often gone by statistics, is divided into literary composition using histogrammic Wave crest and wave trough by text filed
One's own profession, its trough basis for estimation is the angle point number of continuous q row to be less than 1/4 or the 1/3 of angle point meansigma methodss number be considered as trough, 3≤q
≤6.Then the method utilizing angle point density fusion removes the background area of line of text or pseudo- line of text.
Further, the method for described angle point density fusion refers to filter out the background of line of text remaining or removes pseudo-text
One's own profession, its filtering rule utilizes H*1/2H (H is the height of line of text) horizontal sliding window mouth to carry out level slip, removes angle point close
Less than the region of threshold value C, C is angle point number to degree, and the rectangle frame finally again rectangle frame being smaller than H is fused into new text
OK.
Further, the described polarity based on bianry image judges it is the polarity carrying out based on local OTSU bianry image
Judge, first, carry out the binaryzation of line of text using local OTSU method, then four borders after above-mentioned bianry image are made
For sub-pixel point, carry out four connected region seed fill algorithm filling, Filling power is p, 0<p<255, finally calculate black and white two-value
Shared ratio, heavy color is then the polarity of word.
Further, described optimal threshold is adjusted to according to polarity judged result refers to calculate using local OTSU algorithm
Go out threshold value T in each block, but do not carry out binary conversion treatment.Then utilize text polarity judged result, then change present threshold value
For optimal threshold, amended optimal threshold is finally utilized to execute binarization operation.
Further, described modification present threshold value specifically refers to for optimal threshold, if polarity judges text color being
Black, then new threshold value T1=T-T*0.1, conversely, during white text, new threshold value T1=T+T*0.1.
The invention has the beneficial effects as follows：In the String localization stage, it is basic as word that the present invention chooses healthy and strong angle point
Feature, then completes the text filed positioning of candidate by coarse positioning, and reservation as much as possible is text filed；Recycle fine positioning
Realize segmentation and the verification of text filed line of text.Text extraction stage, is judged and office by the polarity based on bianry image
The method that portion OTSU combines completes optimal threshold under complex background and chooses difficult problem.Through the many experiments to the present invention
Test, the String localization of the present invention and extraction algorithm have good vigorousness to complex background video.
Brief description
Fig. 1 is the inventive method flow chart.
Specific embodiment
Below in conjunction with accompanying drawing, the invention will be further described.
As shown in figure 1, a kind of healthy and strong complex background Video Text Location that provides of the present invention and abstracting method, including with
Lower step：
Pretreatment：Input video frame, frame of video is converted into gray level image, carries out Corner Detection to frame of video, obtains angle
Point two-value scattergram；
Text coarse positioning：First with angle point two-value scattergram, region merging technique is carried out by sliding window, form two-value literary composition
One's respective area scattergram, then filters out the angle point in non-textual region using four attributes of two-value text regional distribution chart, real
Existing text filed coarse positioning；
Text fine positioning：By the method for angle point floor projection and angle point density fusion, realize line of text fine positioning；
Text extracts：The polarity that line of text image after positioning is carried out based on bianry image judges, then using improvement
Local OTSU algorithm calculate threshold value in each piece of region, and optimal threshold is adjusted to according to polarity judged result, finally complete
Become the binaryzation of line of text.
Further, through region, fusion obtains described two-value text regional distribution chart, specifically selects a n*n's
Rectangle frame, 5≤n≤15, centered on angle point, angle point color is set to the color in rectangle frame region, has traveled through all angle points,
Finally give two-value text regional distribution chart.
Further, described two-value text regional distribution chart has multiple connected regions.
Further, the attribute of described two-value text regional distribution chart includes：Area, Saturation, Ratio and
Position.
Further, the described angle point filtering out in non-textual region refers to by four attribute character Area,
Saturation, Ratio and Position are progressively filtered, specifically：
First, Area attribute filter refer to each of current two-value text regional distribution chart connected region Area by
It is ranked up according to size, filter out relatively small region Area；Then, Saturation filters and refers to calculate each even
The ratio of the area A (Rect) of area A (Area) and boundary rectangle frame Rect of logical region Area, Saturation=A
(Area)/A (Rect), (Saturation ∈ (0,1)), filter out the less connected region of Saturation；Then, Ratio
Filter and refer to calculate the depth-width ratio of each connected region boundary rectangle frame, filter out Ratio and be more than 1：2.5 connected region；?
Afterwards, Position filter method specifically calculates the positional information of connected region, filters out the connected region of part in frame of video 2/3
Domain.
Further, the method for described angle point floor projection and angle point density fusion refers to：
First, the angle point rectangular histogram often gone by statistics, is divided into literary composition using histogrammic Wave crest and wave trough by text filed
One's own profession, its trough basis for estimation is the angle point number of continuous q row to be less than 1/4 or the 1/3 of angle point meansigma methodss number be considered as trough, 3≤q
≤6.Then the method utilizing angle point density fusion removes the background area of line of text or pseudo- line of text.
Further, the method for described angle point density fusion refers to filter out the background of line of text remaining or removes pseudo-text
One's own profession, its filtering rule utilizes H*1/2H (H is the height of line of text) horizontal sliding window mouth to carry out level slip, removes angle point close
Less than the region of threshold value C, C is angle point number to degree, and the rectangle frame finally again rectangle frame being smaller than H is fused into new text
OK.
Further, the described polarity based on bianry image judges it is the polarity carrying out based on local OTSU bianry image
Judge, first, carry out the binaryzation of line of text using local OTSU method, then four borders after above-mentioned bianry image are made
For sub-pixel point, carry out four connected region seed fill algorithm filling, Filling power is p, 0<p<255, finally calculate black and white two-value
Shared ratio, heavy color is then the polarity of word.
Further, described optimal threshold is adjusted to according to polarity judged result refers to calculate using local OTSU algorithm
Go out threshold value T in each block, but do not carry out binary conversion treatment.Then utilize text polarity judged result, then change present threshold value
For optimal threshold, amended optimal threshold is finally utilized to execute binarization operation.
Further, described modification present threshold value specifically refers to for optimal threshold, if polarity judges text color being
Black, then new threshold value T1=T-T*0.1, conversely, during white text, new threshold value T1=T+T*0.1.
Embodiment
The realization of the present embodiment, comprises the following steps：
1st, input video frame, carries out pretreatment to frame of video, for example, be converted into gray level image；Using Harris algorithm to regarding
Frequency frame carries out Corner Detection, obtains the angle point two-value scattergram that background is that black, angle point are white；
2nd, the rectangle frame of 6*6, centered on angle point, the face consistent by being set to angle point in the rectangle frame of each angle point are utilized
Color, obtains two-value text regional distribution chart；
3rd, carry out String localization using by the thick text positioning method to essence；
Coarse positioning：Four attributes removals choosing two-value text regional distribution chart are pseudo- text filed, obtain candidate's text area
Domain.It is area, saturation, depth-width ratio and position respectively, be designated as Area, Saturation, Ratio, Position.
Area：Find each connected region Area, then according to the size of connected region is ranked up, filter out
Relatively small region.Because the less region of area is not necessarily the main contents of frame of video reflection, and small area region
It is easy to be filtered.
Saturation：The i.e. saturation feature of angle point, counts the boundary rectangle frame Rect of each connected region.Filter out
The less connected region of Saturation.
Saturation=A (Area)/A (Rect), (Saturation ∈ (0,1))；
Wherein A (Area) is the area of connected domain, and A (Rect) is the area of the boundary rectangle frame of connected domain.Due to video
In overlapping text be all level, the therefore value of Saturation is close to 1, and the Saturation of pseudo- text connected region
It is worth close to 0.
Ratio：Connected region boundary rectangle frame depth-width ratio, according to the own characteristic of Chinese character, it is big that the present invention filters out Ratio
In 1：2.5 connected region.
Position：The positional information of connected region, filters out in frame of video 2/3 connected region.Because in frame of video
Overlapping text information be typically all in the lower section of frame of video.
Fine positioning：Due to multiline text or pseudo- text may be comprised in the connected region of each candidate, therefore, it can
Carry out floor projection using angle point, the connected region of each candidate is divided into line of text and realizes being accurately positioned；Finally utilize
Angle point density fusion method filters further to the background of line of text or pseudo- line of text, and filter window size is that (H is H*1/2H
The height of line of text), the rectangle frame again rectangle frame being smaller than H afterwards is fused into new line of text, and line of text positioning completes；
4th, text extracts.The present invention judges mutually to tie with improved local OTSU binaryzation using the polarity based on bianry image
The method closed completes the extraction of line of text.
Local OTSU binaryzation：Line of text image division such as is at the region of size, the size of each zonule is H*H
(height of line of text).Each region carries out local OTSU binaryzation；
Polarity based on bianry image judges：Using four borders after above-mentioned bianry image as sub-pixel point, carry out
Four connected region seed fill algorithm is filled, and the value of filling is 128, finally calculates the ratio shared by black and white two-value, heavy face
Color is then the polarity of word；
Improved local OTSU binaryzation：First to line of text image procossing as the OTSU algorithm of local, by line of text
It is divided into the region that multiple sizes are H*H, then execution OTSU algorithm calculates threshold value T in each block, but does not now carry out two
Value is processed.Because T now is not the optimal threshold of text segmentation.It is thus desirable to according to text polarity judged result, then repair
Change this threshold value (if polarity judges text color for black, new threshold value T1=T-T*0.1, conversely, during white text, new threshold
Value T1=T+T*0.1), finally utilize amended optimal threshold to execute binarization operation.
Claims (10)
1. a kind of healthy and strong complex background Video Text Location and abstracting method are it is characterised in that comprise the following steps：
Pretreatment：Input video frame, frame of video is converted into gray level image, carries out Corner Detection to frame of video, obtains angle point two
Distribution value figure；
Text coarse positioning：First with angle point two-value scattergram, region merging technique is carried out by sliding window, form two-value text area
Domain scattergram, then filters out the angle point in non-textual region using four attributes of two-value text regional distribution chart, realizes literary composition
One's respective area coarse positioning；
Text fine positioning：By the method for angle point floor projection and angle point density fusion, realize line of text fine positioning；
Text extracts：The polarity that line of text image after positioning is carried out based on bianry image judges, then utilizes improved office
Portion's OTSU algorithm calculates the threshold value in each piece of region, and is adjusted to optimal threshold according to polarity judged result, finally completes literary composition
The binaryzation of one's own profession.
2. method according to claim 1 is it is characterised in that described two-value text regional distribution chart merges through region
Obtain, specifically select the rectangle frame of a n*n, 5≤n≤15, centered on angle point, the color in rectangle frame region is set to
Angle point color, has traveled through all angle points, finally gives two-value text regional distribution chart.
3. method according to claim 2 is it is characterised in that described two-value text regional distribution chart has multiple connections
Region.
4. method according to claim 1 is it is characterised in that the attribute of described two-value text regional distribution chart includes：
Area, Saturation, Ratio and Position.
5. method according to claim 1 is it is characterised in that the described angle point filtering out in non-textual region refers to lead to
Cross four attribute character Area, Saturation, Ratio and Position are progressively filtered, specifically：
First, Area attribute filters and refers to each of current two-value text regional distribution chart connected region Area according to face
Long-pending size is ranked up, and filters out relatively small region Area；Then, Saturation filters and refers to calculate each connected region
The ratio of the area A (Rect) of area A (Area) and boundary rectangle frame Rect of domain Area, Saturation=A (Area)/A
(Rect), (Saturation ∈ (0,1)), filters out the less connected region of Saturation；Then, Ratio filters and refers to
Calculate the depth-width ratio of each connected region boundary rectangle frame, filter out Ratio and be more than 1：2.5 connected region；Finally,
Position filter method specifically calculates the positional information of connected region, filters out the connected region of part in frame of video 2/3.
6. method according to claim 1 is it is characterised in that the side of described angle point floor projection and angle point density fusion
Method refers to：
First, the angle point rectangular histogram often gone by statistics, is divided into line of text using histogrammic Wave crest and wave trough by text filed,
Its trough basis for estimation is the angle point number of continuous q row to be less than 1/4 or the 1/3 of angle point meansigma methodss number be considered as trough, 3≤q≤6.
Then the method utilizing angle point density fusion removes the background area of line of text or pseudo- line of text.
7. method according to claim 6 is it is characterised in that the method for described angle point density fusion refers to filter out literary composition
The background of one's own profession remaining or the pseudo- line of text of removal, its filtering rule utilizes H*1/2H (H is the height of line of text) horizontal sliding window
Mouth carries out level slip, removes the region that angle point density is less than threshold value C, and C is angle point number, is finally smaller than rectangle frame again
The rectangle frame of H is fused into new line of text.
8. method according to claim 1 is it is characterised in that the described polarity based on bianry image judges it is based on office
The polarity that portion's OTSU bianry image is carried out judges, first, carries out the binaryzation of line of text using local OTSU method, then upper
State four borders after bianry image as sub-pixel point, carry out four connected region seed fill algorithm filling, Filling power is p, 0
<p<255, finally calculate the ratio shared by black and white two-value, heavy color is then the polarity of word.
9. method according to claim 1 is it is characterised in that described is adjusted to optimal threshold according to polarity judged result
Refer to calculate threshold value T in each block using local OTSU algorithm, but do not carry out binary conversion treatment.Then utilize text polarity
Judged result, then change present threshold value for optimal threshold, finally utilize amended optimal threshold to execute binarization operation.
10. method according to claim 9 it is characterised in that described modification present threshold value be optimal threshold specifically
Refer to, if polarity judges text color for black, new threshold value T1=T-T*0.1, conversely, during white text, new threshold value T1=T
+T*0.1.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610778073.2A CN106407969A (en) | 2016-08-30 | 2016-08-30 | Robust complex background video text positioning and extracting method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610778073.2A CN106407969A (en) | 2016-08-30 | 2016-08-30 | Robust complex background video text positioning and extracting method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106407969A true CN106407969A (en) | 2017-02-15 |
Family
ID=58003924
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610778073.2A Pending CN106407969A (en) | 2016-08-30 | 2016-08-30 | Robust complex background video text positioning and extracting method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106407969A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107302718A (en) * | 2017-08-17 | 2017-10-27 | 河南科技大学 | A kind of video caption area positioning method based on Corner Detection |
| CN107688788A (en) * | 2017-08-31 | 2018-02-13 | 平安科技（深圳）有限公司 | Document charts abstracting method, electronic equipment and computer-readable recording medium |
| CN109993165A (en) * | 2019-03-28 | 2019-07-09 | 永康市几米电子科技有限公司 | The identification of tablet plate medicine name and tablet plate information acquisition method, device and system |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101593277A (en) * | 2008-05-30 | 2009-12-02 | 电子科技大学 | A kind of complicated color image Chinese version zone automatic positioning method and device |
| CN103268481A (en) * | 2013-05-29 | 2013-08-28 | 焦点科技股份有限公司 | Method for extracting text in complex background image |
| CN104182750A (en) * | 2014-07-14 | 2014-12-03 | 上海交通大学 | Extremum connected domain based Chinese character detection method in natural scene image |
-
2016
- 2016-08-30 CN CN201610778073.2A patent/CN106407969A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101593277A (en) * | 2008-05-30 | 2009-12-02 | 电子科技大学 | A kind of complicated color image Chinese version zone automatic positioning method and device |
| CN103268481A (en) * | 2013-05-29 | 2013-08-28 | 焦点科技股份有限公司 | Method for extracting text in complex background image |
| CN104182750A (en) * | 2014-07-14 | 2014-12-03 | 上海交通大学 | Extremum connected domain based Chinese character detection method in natural scene image |
Non-Patent Citations (4)
| Title |
|---|
| 张洋: "电视视频字幕文字的提取方法研究", 《中国优秀硕士学位论文全文数据库_信息科技辑》 * |
| 王刚: "新闻视频字幕的自动提取和识别", 《中国优秀硕士学位论文全文数据库_信息科技辑》 * |
| 胡倩 等: "自然场景下的文本定位", 《电脑知识与技术》 * |
| 黄晓冬: "基于特征融合的视频文本获取研究", 《中国博士学位论文全文数据库_信息科技辑》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107302718A (en) * | 2017-08-17 | 2017-10-27 | 河南科技大学 | A kind of video caption area positioning method based on Corner Detection |
| CN107302718B (en) * | 2017-08-17 | 2019-12-10 | 河南科技大学 | video subtitle area positioning method based on angular point detection |
| CN107688788A (en) * | 2017-08-31 | 2018-02-13 | 平安科技（深圳）有限公司 | Document charts abstracting method, electronic equipment and computer-readable recording medium |
| CN107688788B (en) * | 2017-08-31 | 2021-01-08 | 平安科技（深圳）有限公司 | Document chart extraction method, electronic device and computer readable storage medium |
| CN109993165A (en) * | 2019-03-28 | 2019-07-09 | 永康市几米电子科技有限公司 | The identification of tablet plate medicine name and tablet plate information acquisition method, device and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104751142B (en) | A kind of natural scene Method for text detection based on stroke feature | |
| CN107679502A (en) | A kind of Population size estimation method based on the segmentation of deep learning image, semantic | |
| CN101183425B (en) | Guangdong and Hong Kong license plate locating method | |
| CN105205488B (en) | Word area detection method based on Harris angle points and stroke width | |
| CN109255350B (en) | New energy license plate detection method based on video monitoring | |
| CN102542268B (en) | Method for detecting and positioning text area in video | |
| Yang et al. | Lecture video indexing and analysis using video ocr technology | |
| CN104244073B (en) | Automatic detecting and recognizing method of scroll captions in videos | |
| CN1312625C (en) | Character extracting method from complecate background color image based on run-length adjacent map | |
| CN104077577A (en) | Trademark detection method based on convolutional neural network | |
| CN101515325A (en) | Character extracting method in digital video based on character segmentation and color cluster | |
| CN101593277A (en) | A kind of complicated color image Chinese version zone automatic positioning method and device | |
| CN104598907B (en) | Lteral data extracting method in a kind of image based on stroke width figure | |
| CN102208023A (en) | Method for recognizing and designing video captions based on edge information and distribution entropy | |
| CN104050684B (en) | A kind of video frequency motion target sorting technique based on on-line training and system | |
| CN103336961A (en) | Interactive natural scene text detection method | |
| CN107977645B (en) | Method and device for generating video news poster graph | |
| CN104766076A (en) | Detection method and device for video images and texts | |
| CN104299009A (en) | Plate number character recognition method based on multi-feature fusion | |
| CN101106716A (en) | A shed image division processing method | |
| KR20120019425A (en) | Image processing device, method, and program | |
| CN106407969A (en) | Robust complex background video text positioning and extracting method | |
| CN113191358B (en) | Metal part surface text detection method and system | |
| CN102073872A (en) | Image-based method for identifying shape of parasite egg | |
| CN106295627A (en) | For identifying the method and device of word psoriasis picture |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170215 |
|
| RJ01 | Rejection of invention patent application after publication |