CN105608470A - Image retrieval method based on Weibull distribution statistical model - Google Patents
Image retrieval method based on Weibull distribution statistical model Download PDFInfo
- Publication number
- CN105608470A CN105608470A CN201510995574.1A CN201510995574A CN105608470A CN 105608470 A CN105608470 A CN 105608470A CN 201510995574 A CN201510995574 A CN 201510995574A CN 105608470 A CN105608470 A CN 105608470A
- Authority
- CN
- China
- Prior art keywords
- image
- weibull distribution
- similarity
- images
- parameter
- 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
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- 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
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
Abstract
The present invention discloses an image retrieval method based on a Weibull distribution statistical model. According to the method, in the process of image retrieval, non-downsampling contourlet transform is used to decompose an image, a Weibull distribution probability density function is used to take the shape parameter and size parameter of the image obtained by decomposition as image characteristics, an image characteristic library is for retrieval, then the above operations are carried out on an image to be retrieved, finally by using a calculation method of using a Euclidean distance as the similarity between images, the similarities between the images is calculated, and results are sorted and outputted according to the magnitude of the similarities. A test result shows that the system can effectively characterize the various features of the images, and the retrieval efficiency is greatly improved. In addition, the method has the characteristics of simple design, easy implementation, and no need of manual tagging in an image characteristic extraction process, a tedious and error-prone manual tagging process is avoided, and the usability of an image retrieval system is enhanced.
Description
Technical field
The invention belongs to image processing field, especially a kind of improve recall precision based on Weibull distribution statistical modelImage search method.
Background technology
Image retrieval technologies (TBIR) has in the past been continued to use traditional text retrieval technique, and it does not consider that image itself is intrinsicThe content characteristics such as color, texture, shape, but describe and retrieving images with keyword. There is following defect: to databaseIn each sub-picture carry out artificial mark and need to expend a large amount of time and manpower; Use keyword to be difficult to accurate description content thousandDiffer from the intension of ten thousand other images; Manually choose keyword and can comprise strong subjectivity, may cause the deviation on image understanding, straightConnect the retrieval effectiveness that affects image. In order to overcome above-mentioned defect, CBIR technology (CBIR) is arisen at the historic moment, shouldRetrieval technique comprises that Image Visual Feature is extracted and characteristic similarity calculates two links, the speed that can effectively improve retrieval withEfficiency, has better objectivity and versatility, is more suitable for the retrieval of extensive image library.
In Image Visual Feature leaching process, especially for the image that contains a large amount of texture informations, how accurately tableLevy textural characteristics and just become the deciding factor that affects retrieval effectiveness. Existing image texture characteristic extracting method, as FourierConversion, wavelet transformation, multiple dimensioned geometric transformation etc. all respectively have its limitation in this regard, cause similarity long computing time, shadowRing recall precision.
Summary of the invention
The present invention is in order to solve the existing above-mentioned technical problem of prior art, and a kind of recall precision that improves is providedBased on the image search method of Weibull distribution statistical model.
Technical solution of the present invention is: a kind of image search method based on Weibull distribution statistical model, its featureBe to carry out in accordance with the following steps:
Step 1: utilize non-down sampling contourlet transform to decompose image, obtain multiple dimensioned multidirectional several high frequencyBand;
Step 2: utilize Weibull distribution probability density function to carry out statistical modeling to each width picture breakdown gained high-frequency sub-band, estimateCount out form parameter p and scale parameter c, and spy using obtained form parameter p and scale parameter c as every piece imageLevy, as characteristics of image storehouse with use to be retrieved;
Step 3: treat retrieving images I and process operation;
Step 4: utilize the computational methods of Euclidean distance as similarity between image, calculate the similarity between different images,From big to small result is sorted, exported according to similarity.
Described step 2 is as follows:
Step 21: definition Weibull distribution statistical model expression formula is:
And have 0<x<∞, p>0, c>0;
Step 22: for two parameter Estimation of Weibull distribution probability density function, establishTo obey weberThe random sample distributing, form parameter p can be calculated by log function:
Can solve above-mentioned nonlinear equation by Newton iteration method, the computational methods of scale parameter c are:
Obtain two parameters of Weibull distribution probability density function, respectively to several high frequencies that obtain after each width picture breakdownSubband calculates form parameter p and scale parameter c, the feature using the value obtaining as every piece image, as characteristics of image storehouseWith use to be retrieved.
Described step 3 is as follows:
Step 31: input image I to be retrieved, utilize non-down sampling contourlet transform to treat retrieving images I and decompose, obtain manyMultidirectional several high-frequency sub-band of yardstick;
Step 33: according to the method for step 2, calculate form parameter p and the scale parameter c of image I to be retrieved, obtain figure to be retrievedThe feature of picture I。
Described step 4 is as follows:
Step 41: adopt Euclidean distance to calculate characteristics of image to be retrievedWith the feature in characteristics of image storehouseSimilarity,For the distance between image J in retrieving images I and storehouse,I the component place of characteristic vector of expression image ICharacteristic value, the Euclidean distance of two width images is defined as:
Step 42: willGet reciprocally as the similarity between two width images, before all similarities are got inverse, first add and beNumber, get, result for retrieval is arranged to output from big to small according to similarity, obtain result for retrieval.
Utilization of the present invention has the non-down sampling contourlet transform exploded view picture of anisotropy and translation invariance, then knotClose and use the character that Weibull distribution statistical model can rendering image textural characteristics, obtain the characteristic vector of rendering image, designAn image search method based on Weibull distribution statistical model. Owing to by non-down sampling contourlet transform, image being carried outDecompose, can obtain having the subband of anisotropy and translation invariance; The Weibull distribution statistical model of two ginsengs is by the shape of trying to achieveShape parameter and scale parameter are determined shape jointly, have solved the free degree and poor the asking of flexibility ratio in the time of data of description statistical informationTopic, has reduced time complexity. Experimental result shows, this system is each feature of token image effectively, improves greatlyRecall precision. In addition, the method has simplicity of design, is easy to realize, image characteristics extraction process is without spies such as artificial marksPoint, has avoided the loaded down with trivial details and artificial mark process of fallibility, has strengthened the practicality of image indexing system.
Detailed description of the invention
Method of the present invention comprises the following steps:
Step 1: utilize non-down sampling contourlet transform to decompose image, obtain multiple dimensioned multidirectional several high frequencyBand;
Step 2: utilize Weibull distribution probability density function to carry out statistical modeling to each width picture breakdown gained high-frequency sub-band, estimateCount out form parameter p and scale parameter c, and spy using obtained form parameter p and scale parameter c as every piece imageLevy, as characteristics of image storehouse with use to be retrieved; Concrete steps are as follows:
Step 21: definition Weibull distribution statistical model expression formula is:
And have 0<x<∞, p>0, c>0; Image is after non-down sampling contourlet transform, and the statistical distribution of high-frequency sub-band coefficient accords withClose the Weibull distribution statistical model that parameter is (c, p);
Step 22: for two parameter Estimation of Weibull distribution probability density function, establishTo obey weberThe random sample distributing, form parameter p can be calculated by log function:
Can solve above-mentioned nonlinear equation by Newton iteration method, the computational methods of scale parameter c are:
Obtain two parameters of Weibull distribution probability density function, respectively to several high frequencies that obtain after each width picture breakdownSubband calculates form parameter p and scale parameter c, the feature using the value obtaining as every piece image, as characteristics of image storehouseWith use to be retrieved.
Step 3: treat retrieving images I and process operation; Concrete steps are as follows:
Step 31: input image I to be retrieved, utilize non-down sampling contourlet transform to treat retrieving images I and decompose, obtain manyMultidirectional several high-frequency sub-band of yardstick;
Step 33: according to the method for step 2, calculate form parameter p and the scale parameter c of image I to be retrieved, obtain figure to be retrievedThe feature of picture I。
Step 4: utilize the computational methods of Euclidean distance as similarity between image, calculate similar between different imagesDegree, sorts, exports result from big to small according to similarity. Concrete steps are as follows:
Step 41: adopt Euclidean distance to calculate characteristics of image to be retrievedWith the feature in characteristics of image storehouseSimilarity,For the distance between image J in retrieving images I and storehouse,I the component place of characteristic vector of expression image ICharacteristic value, the Euclidean distance of two width images is defined as:
Step 42: willGet reciprocal as the similarity between two width images, wherein, for fear of institute's query image withThis image distance in image library is the 1/0 meaningless result occurring, before all similarities are got inverse, first add one infiniteThe little coefficient that can not cause that sequence changes, get, result for retrieval is arranged to output from big to small according to similarity,To result for retrieval.
The contrast of table 1 feature extraction time
Contrast computing time of table 2 similarity
The contrast of table 3 average precision
Documents: N.E.Lasmar, Y.Berthoumieu.GaussianCopulaMultivariateModelingforTextureImageRetrievalUsingWaveletTransforms.IEEETrans.onImageProcessing,2014,23(5):2246-2261。
Above table 1, table 2 and table 3, be respectively documents and the present invention to four image library VISTeX, Brodatz,Image in ALOT and STeX is retrieved the comparison of the data that obtain. Wherein, table 1 is carried for documents and feature of the present inventionGet the contrast of time span, table 2 is the contrast of documents and similarity length computing time of the present invention, and table 3 is contrast literary compositionOffer the average precision contrast of retrieving with the present invention. As can be seen from the comparison result, the present invention is in time complexity and inspectionOn Suo Xiaoguo, there is greater advantage.
Claims (4)
1. the image search method based on Weibull distribution statistical model, is characterized in that carrying out in accordance with the following steps:
Step 1: utilize non-down sampling contourlet transform to decompose image, obtain multiple dimensioned multidirectional several high frequencyBand;
Step 2: utilize Weibull distribution probability density function to carry out statistical modeling to each width picture breakdown gained high-frequency sub-band, estimateCount out form parameter p and scale parameter c, and spy using obtained form parameter p and scale parameter c as every piece imageLevy, as characteristics of image storehouse with use to be retrieved;
Step 3: treat retrieving images I and process operation;
Step 4: utilize the computational methods of Euclidean distance as similarity between image, calculate the similarity between different images,From big to small result is sorted, exported according to similarity.
2. the image search method based on Weibull distribution statistical model according to claim 1, is characterized in that described stepRapid 2 as follows:
Step 21: definition Weibull distribution statistical model expression formula is:
And have 0<x<∞, p>0, c>0;
Step 22: for two parameter Estimation of Weibull distribution probability density function, establishTo obey weber pointThe random sample of cloth, form parameter p can be calculated by log function:
Can solve above-mentioned nonlinear equation by Newton iteration method, the computational methods of scale parameter c are:
Obtain two parameters of Weibull distribution probability density function, respectively to several high frequencies that obtain after each width picture breakdownSubband calculates form parameter p and scale parameter c, the spy using the value obtaining as every piece image
Levy, as characteristics of image storehouse with use to be retrieved.
3. the image search method based on Weibull distribution statistical model according to claim 2, is characterized in that described stepRapid 3 as follows:
Step 31: input image I to be retrieved, utilize non-down sampling contourlet transform to treat retrieving images I and decompose, obtain manyMultidirectional several high-frequency sub-band of yardstick;
Step 33: according to the method for step 2, calculate form parameter p and the scale parameter c of image I to be retrieved, obtain figure to be retrievedThe feature of picture I。
4. the image search method based on Weibull distribution statistical model according to claim 3, is characterized in that described stepRapid 4 as follows:
Step 41: adopt Euclidean distance to calculate characteristics of image to be retrievedWith the feature in characteristics of image storehouseSimilarity,For the distance between image J in retrieving images I and storehouse,I the component place of characteristic vector of expression image ICharacteristic value, the Euclidean distance of two width images is defined as:
Step 42: willGet reciprocally as the similarity between two width images, before all similarities are got inverse, first add and beNumber, get, result for retrieval is arranged to output from big to small according to similarity, obtain result for retrieval.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510995574.1A CN105608470A (en) | 2015-12-28 | 2015-12-28 | Image retrieval method based on Weibull distribution statistical model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510995574.1A CN105608470A (en) | 2015-12-28 | 2015-12-28 | Image retrieval method based on Weibull distribution statistical model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105608470A true CN105608470A (en) | 2016-05-25 |
Family
ID=55988394
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510995574.1A Pending CN105608470A (en) | 2015-12-28 | 2015-12-28 | Image retrieval method based on Weibull distribution statistical model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105608470A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106682157A (en) * | 2016-12-24 | 2017-05-17 | 辽宁师范大学 | Image retrieval method based on Weibull distribution parameters |
| CN106815314A (en) * | 2016-12-22 | 2017-06-09 | 辽宁师范大学 | Image search method based on amplitude phase hybrid modeling |
| CN110175655A (en) * | 2019-06-03 | 2019-08-27 | 中国科学技术大学 | Data identification method and device, storage medium and electronic equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102073995A (en) * | 2010-12-30 | 2011-05-25 | 上海交通大学 | Color constancy method based on texture pyramid and regularized local regression |
| US8879800B2 (en) * | 2011-06-15 | 2014-11-04 | Honeywell International Inc. | Quality driven image processing for ocular recognition system |
| CN104199923A (en) * | 2014-09-01 | 2014-12-10 | 中国科学院自动化研究所 | Massive image library retrieving method based on optimal K mean value Hash algorithm |
| CN104484432A (en) * | 2014-12-20 | 2015-04-01 | 辽宁师范大学 | Color image searching method based on quaternion exponential moment |
-
2015
- 2015-12-28 CN CN201510995574.1A patent/CN105608470A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102073995A (en) * | 2010-12-30 | 2011-05-25 | 上海交通大学 | Color constancy method based on texture pyramid and regularized local regression |
| US8879800B2 (en) * | 2011-06-15 | 2014-11-04 | Honeywell International Inc. | Quality driven image processing for ocular recognition system |
| CN104199923A (en) * | 2014-09-01 | 2014-12-10 | 中国科学院自动化研究所 | Massive image library retrieving method based on optimal K mean value Hash algorithm |
| CN104484432A (en) * | 2014-12-20 | 2015-04-01 | 辽宁师范大学 | Color image searching method based on quaternion exponential moment |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106815314A (en) * | 2016-12-22 | 2017-06-09 | 辽宁师范大学 | Image search method based on amplitude phase hybrid modeling |
| CN106682157A (en) * | 2016-12-24 | 2017-05-17 | 辽宁师范大学 | Image retrieval method based on Weibull distribution parameters |
| CN110175655A (en) * | 2019-06-03 | 2019-08-27 | 中国科学技术大学 | Data identification method and device, storage medium and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Arivazhagan et al. | Texture classification using ridgelet transform | |
| Varish et al. | Content based image retrieval using statistical features of color histogram | |
| Singha et al. | Content-based image retrieval using the combination of the fast wavelet transformation and the colour histogram | |
| Ma | Content-based image retrieval with HSV color space and texture features | |
| CN105608470A (en) | Image retrieval method based on Weibull distribution statistical model | |
| CN104504406B (en) | A kind of approximate multiimage matching process rapidly and efficiently | |
| CN104484425A (en) | Color image searching method based on multiple features | |
| CN104484432A (en) | Color image searching method based on quaternion exponential moment | |
| Varughese et al. | Content Based Image Retrieval Technique on Texture and Shape Analysis using Wavelet Feature and Clustering Model | |
| CN105426543A (en) | Image retrieval method based on Bessel statistic model | |
| Yakin et al. | Application of content based image retrieval in digital image search system | |
| CN105631001A (en) | Multi-feature fusion based color image retrieval method for HSV space image retrieval | |
| Babu et al. | Content based image retrieval using color, texture, shape and active re-ranking method | |
| CN104504049A (en) | Retrieval method of color images based on quaternion Harmonic-Fourier moments | |
| Thepade et al. | Image cataloging using Bayes, Function, Lazy, Rule, Tree classifier families with row mean of Fourier transformed image content | |
| CN114565772A (en) | Set feature extraction method and device, electronic equipment and storage medium | |
| CN106815314A (en) | Image search method based on amplitude phase hybrid modeling | |
| Youssef et al. | An efficient content-based image retrieval system integrating wavelet-based image sub-blocks with dominant colors and texture analysis | |
| Yu et al. | Iterative Recognition of Bird's Nest in Aerial Photograph of High Voltage Transmission Tower | |
| CN106682157A (en) | Image retrieval method based on Weibull distribution parameters | |
| Wang et al. | Digital image retrieval technology based on multi-feature | |
| Shambharkar et al. | A comparative study on retrieved images by content based image retrieval system based on binary tree, color, texture and canny edge detection approach | |
| Arivazhagan et al. | Texture classification using curvelet transform | |
| CN107992612A (en) | Texture images retrieval based on truncation broad sense Cauchy modeling | |
| Gancarczyk | Decision tree based approach to craquelure identification in old paintings |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20160525 |
|
| WD01 | Invention patent application deemed withdrawn after publication |