CN105550706A - Method of recognizing 2D image and 3D image - Google Patents
Method of recognizing 2D image and 3D image Download PDFInfo
- Publication number
- CN105550706A CN105550706A CN201510924850.5A CN201510924850A CN105550706A CN 105550706 A CN105550706 A CN 105550706A CN 201510924850 A CN201510924850 A CN 201510924850A CN 105550706 A CN105550706 A CN 105550706A
- Authority
- CN
- China
- Prior art keywords
- image
- sample
- right half
- correlogram
- identified
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
Abstract
The invention relates to a method of recognizing a 2D image and a 3D image. The method comprises the following steps: (1) multiple 2D images and multiple 3D images serve as a sample image set; (2) each image in the sample image set is divided into a left half image sample and a right half image sample with the same size, and color autocorrelogram features of the left half image sample and the right half image sample are extracted respectively; (3) the difference C1 between the color autocorrelogram features of the left half image sample and the right half image sample and a correlation coefficient C2 between the left half image sample and the right half image sample are calculated, and the C1 and the C2 serve as evaluation indexes; (4) the C1 and the C2 are integrated, a decision-making model is built, and the optimal classification boundary is obtained; and (5) through extracting the color autocorrelogram features and the correlation coefficient features of the left half image and the right half image of the to-be-recognized image, the to-be-recognized image is classified and marked based on the optimal classification boundary.
Description
Technical field
The present invention relates to a kind of recognition methods, particularly relate to the recognition methods of a kind of 2D image and 3D rendering.
Background technology
3D rendering, 3D video are more and more welcomed by the general public due to the stereo display effect of its uniqueness.Existing various picture browsing equipment often can not identify 2D image and 3D rendering automatically, and needs user to carry out manual switchover, causes image browsing inconvenience.
Summary of the invention
In view of this, necessaryly provide a kind of and can be applicable to the automatic identification 2D image of various picture browsing equipment and the method for 3D rendering.
The invention provides the recognition methods of a kind of 2D image and 3D rendering, it comprises the following steps:
(1) using several 2D images and several 3D renderings as sample graph image set;
(2) the every piece image in described sample image set is divided sized by an identical left side half figure sample and right half figure sample, and respectively to described left half figure sample and right half figure sample extraction color auto-correlogram feature;
(3) the otherness C between the color auto-correlogram feature calculating the right half figure sample of color auto-correlogram characteristic sum of described left half figure sample
1, and described left half figure sample, related coefficient C between right half figure sample
2, and by C
1, C
2as evaluation index;
(4) by evaluation index C
1and C
2comprehensive and build decision model, obtain optimal classification border; And
(5) by extracting the Zuo Bantu of image to be identified and the color auto-correlogram feature of right half figure and correlation coefficient eigenvalue, with optimal classification border for according to classifying to this image to be identified and mark.
Compared with prior art, the recognition methods of 2D image provided by the invention and 3D rendering has the following advantages: by carrying out the color auto-correlogram signature analysis of left and right half figure sample to several 2D images and 3D rendering, the otherness C between the color auto-correlogram calculating left and right half figure sample
1, and analyze the correlation coefficient eigenvalue C of left and right half figure sample further
2, with C
1and C
2build decision model and obtain optimal classification border.Treat recognition image as a reference with this optimal classification border to classify.The method can realize the automatic identification to 2D image and 3D rendering, and degree of accuracy is higher, simple and convenient, substantially reduce the number the manual switchover of user, for image browsing is provided convenience.The method can be applicable in various picture browsing equipment.
Accompanying drawing explanation
Fig. 1 is the process flow diagram of the recognition methods of 2D image of the present invention and 3D rendering.
Fig. 2 to Fig. 5 be the present embodiment sample image concentrate 2D image and 3D rendering.
Fig. 6 to Fig. 9 be respectively the color auto-correlogram of the left and right half figure sample of Fig. 2 to Fig. 5 histogram (wherein Fig. 6 correspond to Fig. 2, Fig. 7 corresponds to Fig. 3, Fig. 8 corresponds to Fig. 4, Fig. 9 corresponds to Fig. 5, in Fig. 6 to Fig. 9, " left side " represents left half figure sample, and " right side " represents right half figure sample).
Following specific embodiment will further illustrate the present invention in conjunction with above-mentioned accompanying drawing.
Embodiment
Be described further to the recognition methods of 2D image provided by the invention and 3D rendering below.
The invention provides the recognition methods of a kind of 2D image and 3D rendering.The method comprises the following steps:
S1, using several 2D images and several 3D renderings as sample graph image set;
S2, an identical left side half figure sample and right half figure sample sized by being divided by the every piece image in described sample image set, respectively to described left half figure sample and right half figure sample extraction color auto-correlogram feature;
S3, the otherness C between the color auto-correlogram feature calculating the right half figure sample of color auto-correlogram characteristic sum of described left half figure sample
1, calculate the related coefficient C between described left half figure sample, right half figure sample simultaneously
2, and by C
1, C
2as evaluation index;
S4, by evaluation index C
1and C
2comprehensive and build decision model, obtain optimal classification border; And
S5, by color auto-correlogram feature and the correlation coefficient eigenvalue of the Zuo Bantu and right half figure that extract image to be identified, with optimal classification border for according to classifying to this image to be identified and mark.
In step sl, using this sample graph image set as research object, to obtain the C of multiple image
1and C
2, thus better build decision model.
After step S1, the every piece image also comprised before step S2 in sample image set described in a pair carries out the step of down-sampled process.Specifically, by image drop samplings all in described sample image set to fixed size.
In step s 2, relative to common 2D image, the obvious difference of 3D rendering of left-right format is, Zuo Bantu has identical content with right half figure, just there is certain parallax, therefore this method utilizes this characteristic of 3D rendering to carry out feature extraction, thus realizes automatically identifying 2D image and 3D rendering.
Described color auto-correlogram feature refers to the spatial relationship utilizing same color right, color and spatial information is combined and is described image.Concrete, being divided into left and right two parts by obtaining image after down-sampled according to center line, being called left half figure sample, right half figure sample.Left and right half pattern is originally described by color auto-correlogram feature respectively, similarity more between the two.
In step s3, available Euclidean distance weighs the otherness C between the color auto-correlogram feature of the right half figure sample of color auto-correlogram characteristic sum of described left half figure sample
1, i.e. distance between the color auto-correlogram feature of the right half figure sample of the color auto-correlogram characteristic sum of described left half figure sample is C
1.By C
1, C
2reason jointly as evaluation index is: C
1with C
2complementary effect can be played, prevent 2D image (as Fig. 3) similar in content and color having symmetric 2D image (as Fig. 4) and interference is produced to identification.
In the present embodiment, for Fig. 2 to Fig. 5, the C calculated
1and C
2be worth as shown in table 1 below:
Table 1
| Legend | Fig. 2 | Fig. 3 | Fig. 4 | Fig. 5 |
| C 1 | 0.9313 | 0.8166 | 0.2047 | 0.1117 |
| C 2 | 0.3402 | 0.7767 | 0.3047 | 0.8673 |
Fig. 6 to Fig. 9 sets forth the histogram of a left side half figure sample of Fig. 2 to Fig. 5, the color auto-correlogram of right half figure sample.Between left and right half figure sample color auto-correlogram feature (as Fig. 7) of wherein Fig. 4, there is obvious similarity, but the related coefficient of its left and right half figure sample is only 0.3047 (see table 1); And the related coefficient between the left and right half figure sample of Fig. 3 is 0.7767, but by left and right half figure sample color auto-correlogram feature, its not left-right format image can be distinguished.This illustrates C
1, C
2there is certain robustness (robustness) as this method of evaluation index to having interfering image simultaneously.
From table 1, Fig. 6 to Fig. 9, the C described in this method
1and C
2between really serve complementary effect, by C
1and C
2effectively can ensure the accuracy of Images Classification as evaluation index simultaneously.
In step s 4 which, by evaluation index C
1and C
2comprehensive and build decision model and be specially: by evaluation index C
1and C
2combine merga pass decision-tree model to build described decision model or build described decision model by support vector machine, artificial neural network or bayes method.For decision-tree model, by evaluation index C
1, C
2and corresponding category attribute input decision-tree model, optimal classification border can be obtained.This decision-tree model is stored among picture browsing equipment.
In the present embodiment, the optimal classification border obtained is C
1be less than 0.31, C
2be greater than 0.485, identifiable design goes out 3D rendering, and measuring accuracy reaches 95.24%.Through cross validation, discrimination is all more than 90%.
In step s 5, with optimal classification border for according to marking all images to be identified under treating browse through folders in picture browsing equipment.By the decision model be stored in picture browsing equipment, automatic mark is carried out to this image to be identified.Concrete, mark result is stored as a part for the image information of current image to be identified; Treat browse through folders to carry out screening to realize classifying to unlabelled image and marking, to avoid repeating label; The mark of final realization all images under treating browse through folders.
In order to tackle the situation of identification error, the step of carrying out correcting when a pair image to be identified marks by mistake also can be comprised after step s 5.The concrete image for error flag, it shows with the browsing mode of mistake when browsing, and user can correct the mark of present image voluntarily, now only need by the interface opening of mark to user.
Compared with prior art, the recognition methods of 2D image provided by the invention and 3D rendering has the following advantages: by carrying out the color auto-correlogram signature analysis of left and right half figure sample to several 2D images and 3D rendering, the otherness C between the color auto-correlogram calculating left and right half figure sample
1, and analyze the correlation coefficient eigenvalue C of left and right half figure sample further
2, with C
1and C
2build decision model and obtain optimal classification border.Treat recognition image as a reference with this optimal classification border to classify.The method can realize the automatic identification to 2D image and 3D rendering, and degree of accuracy is higher, simple and convenient, substantially reduce the number the manual switchover of user, for image browsing is provided convenience, thus ensure that user obtains higher experience satisfaction.
The explanation of above embodiment just understands method of the present invention and core concept thereof for helping.It should be pointed out that for those skilled in the art, under the premise without departing from the principles of the invention, can also carry out some improvement and modification to the present invention, these improve and modify and also fall in the protection domain of the claims in the present invention.
To the above-mentioned explanation of the disclosed embodiments, professional and technical personnel in the field are realized or uses the present invention.To be apparent for those skilled in the art to the multiple amendment of these embodiments, General Principle as defined herein can without departing from the spirit or scope of the present invention, realize in other embodiments.Therefore, the present invention can not be restricted to these embodiments shown in this article, but will meet the widest scope consistent with principle disclosed herein and features of novelty.
Claims (5)
1. a recognition methods for 2D image and 3D rendering, it comprises the following steps:
(1) using several 2D images and several 3D renderings as sample graph image set;
(2) the every piece image in described sample image set is divided sized by an identical left side half figure sample and right half figure sample, respectively to described left half figure sample and right half figure sample extraction color auto-correlogram feature;
(3) the otherness C between the color auto-correlogram feature calculating the right half figure sample of color auto-correlogram characteristic sum of described left half figure sample
1, and described left half figure sample, related coefficient C between right half figure sample
2, and by C
1, C
2as evaluation index;
(4) by evaluation index C
1and C
2comprehensive and build decision model, obtain optimal classification border; And
(5) by extracting the Zuo Bantu of image to be identified and the color auto-correlogram feature of right half figure and correlation coefficient eigenvalue, with optimal classification border for according to classifying to this image to be identified and mark.
2. recognition methods as claimed in claim 1, is characterized in that, after step (1), before step (2), the every piece image also comprised in sample image set described in a pair carries out the step of down-sampled process.
3. recognition methods as claimed in claim 1, is characterized in that, by evaluation index C in step (4)
1and C
2comprehensive and build decision model and be specially: by evaluation index C
1and C
2combine merga pass decision-tree model to build described decision model or build described decision model by support vector machine, artificial neural network or bayes method.
4. recognition methods as claimed in claim 1, it is characterized in that, in step (5) when image to be identified is arranged in until browse through folders, with optimal classification border for according to classify to this image to be identified and this mark result is stored after mark, with realization unlabelled image to be classified by the screening for the treatment of browse through folders and mark.
5. recognition methods as claimed in claim 1, is characterized in that, also comprises the step of carrying out correcting when a pair image to be identified marks by mistake after step (5).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510924850.5A CN105550706A (en) | 2015-12-13 | 2015-12-13 | Method of recognizing 2D image and 3D image |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510924850.5A CN105550706A (en) | 2015-12-13 | 2015-12-13 | Method of recognizing 2D image and 3D image |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105550706A true CN105550706A (en) | 2016-05-04 |
Family
ID=55829889
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510924850.5A Pending CN105550706A (en) | 2015-12-13 | 2015-12-13 | Method of recognizing 2D image and 3D image |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105550706A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107766871A (en) * | 2017-08-29 | 2018-03-06 | 深圳依偎控股有限公司 | A kind of method and system of Intelligent Recognition 3D pictures |
| CN108133210A (en) * | 2017-12-12 | 2018-06-08 | 上海玮舟微电子科技有限公司 | A kind of picture format recognition methods and device |
| CN109189735A (en) * | 2018-10-23 | 2019-01-11 | 维沃移动通信有限公司 | A kind of preview image displaying method, mobile terminal |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101551823A (en) * | 2009-04-20 | 2009-10-07 | 浙江师范大学 | Comprehensive multi-feature image retrieval method |
| CN103246895A (en) * | 2013-05-15 | 2013-08-14 | 中国科学院自动化研究所 | Image classifying method based on depth information |
| CN103440494A (en) * | 2013-07-04 | 2013-12-11 | 中国科学院自动化研究所 | Horrible image identification method and system based on visual significance analyses |
| CN103955462A (en) * | 2014-03-21 | 2014-07-30 | 南京邮电大学 | Image marking method based on multi-view and semi-supervised learning mechanism |
-
2015
- 2015-12-13 CN CN201510924850.5A patent/CN105550706A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101551823A (en) * | 2009-04-20 | 2009-10-07 | 浙江师范大学 | Comprehensive multi-feature image retrieval method |
| CN103246895A (en) * | 2013-05-15 | 2013-08-14 | 中国科学院自动化研究所 | Image classifying method based on depth information |
| CN103440494A (en) * | 2013-07-04 | 2013-12-11 | 中国科学院自动化研究所 | Horrible image identification method and system based on visual significance analyses |
| CN103955462A (en) * | 2014-03-21 | 2014-07-30 | 南京邮电大学 | Image marking method based on multi-view and semi-supervised learning mechanism |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107766871A (en) * | 2017-08-29 | 2018-03-06 | 深圳依偎控股有限公司 | A kind of method and system of Intelligent Recognition 3D pictures |
| CN108133210A (en) * | 2017-12-12 | 2018-06-08 | 上海玮舟微电子科技有限公司 | A kind of picture format recognition methods and device |
| CN108133210B (en) * | 2017-12-12 | 2022-04-01 | 张家港康得新光电材料有限公司 | Image format identification method and device |
| CN109189735A (en) * | 2018-10-23 | 2019-01-11 | 维沃移动通信有限公司 | A kind of preview image displaying method, mobile terminal |
| CN109189735B (en) * | 2018-10-23 | 2020-09-01 | 维沃移动通信有限公司 | Preview image display method and mobile terminal |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103473551A (en) | Station logo recognition method and system based on SIFT operators | |
| CN105184238A (en) | Human face recognition method and system | |
| CN108629319B (en) | Image detection method and system | |
| CN103258037A (en) | Trademark identification searching method for multiple combined contents | |
| CN103456013B (en) | A kind of method representing similarity between super-pixel and tolerance super-pixel | |
| CN105574063A (en) | Image retrieval method based on visual saliency | |
| CN102968637A (en) | Complicated background image and character division method | |
| CN104850850A (en) | Binocular stereoscopic vision image feature extraction method combining shape and color | |
| CN103778409A (en) | Human face identification method based on human face characteristic data mining and device | |
| CN108197644A (en) | A kind of image-recognizing method and device | |
| CN109740572A (en) | A kind of human face in-vivo detection method based on partial color textural characteristics | |
| CN104123529A (en) | Human hand detection method and system thereof | |
| CN103473571A (en) | Human detection method | |
| CN110728302A (en) | Method for identifying color textile fabric tissue based on HSV (hue, saturation, value) and Lab (Lab) color spaces | |
| CN105718552A (en) | Clothing freehand sketch based clothing image retrieval method | |
| CN106897681A (en) | A kind of remote sensing images comparative analysis method and system | |
| CN104299009A (en) | Plate number character recognition method based on multi-feature fusion | |
| WO2021000829A1 (en) | Multi-dimensional identity information identification method and apparatus, computer device and storage medium | |
| CN101493887A (en) | Eyebrow image segmentation method based on semi-supervision learning and Hash index | |
| CN104156413A (en) | Trademark density based personalized trademark matching recognition method | |
| CN107169417A (en) | Strengthened based on multinuclear and the RGBD images of conspicuousness fusion cooperate with conspicuousness detection method | |
| CN107870992A (en) | Editable image of clothing searching method based on multichannel topic model | |
| CN108875744A (en) | Multi-oriented text lines detection method based on rectangle frame coordinate transform | |
| CN104063701B (en) | Fast electric television stations TV station symbol recognition system and its implementation based on SURF words trees and template matches | |
| CN105550706A (en) | Method of recognizing 2D image and 3D image |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| AD01 | Patent right deemed abandoned |
Effective date of abandoning: 20190716 |
|
| AD01 | Patent right deemed abandoned |