Font Size: a A A

Research On Image Retrieval Algorithm Based On Curvelet

Posted on:2017-03-02Degree:MasterType:Thesis
Country:ChinaCandidate:J R HeFull Text:PDF
GTID:2358330482999439Subject:Instrument Science and Technology
Abstract/Summary:PDF Full Text Request
With the rapid development of computer multimedia technology, digital image capacity is increasing sharply. But these images are disorderly distributed all over the world, the useful information contained in the image can not be used and accessed effectively. At the same time, the images are used more and more by all walks of life, so how the user find the image we need from the vast amounts of images is becoming an increasingly prominent issue. At this point we need a fast and accurate technology to search and access image——image retrieval. Content-based image retrieval (CBIR) technique is a mainstream direction in the field of image retrieval research. Feature extraction is the core content of the CBIR. Image transform domain characteristics'research is the key content of the feature extraction research. Wavelet transform is effective to image singularity detection, so it gets the application and plays its good effect in the field of image retrieval, but its express has its limitations to the image edge information. Curvelet transform can express image edge information well. More abundant and more effective image texture information can be described. Currently although single Curvelet transform used for image feature extraction has its progressive compared with Wavelet transform, but there are still some shortcomings. This paper studies on image retrieval algorithm based on Curvelet transform, the main research work is as follows:1. Considering that the Wavelet transform and Curvelet transform have their own limitations, a feature extraction method is put forward by this paper which combines Curvelet transform with Wavelet transform, and realizes that the image features are expressed more effectively. In the Brodatz texture image library, It expresses the image texture information, compares it with other three algorithms and proves that it is progressive. The algorithm is applied in Corel color images, combining with color histogram features for image retrieval to verify the algorithm's retrieval effect in this image library.2. According to the advantages and disadvantages of the image texture transform domain characteristics and the statistical characteristics, a image retrieval algorithm is put forward by this paper which makes Curvelet?LBP and GLCM to fuse together. The algorithm is applied to the Brodatz texture image library, compares it with other four algorithms and proves that it is progressive. At the same time, in Corel color image repository, It fuses with the color histogram characteristics of the image to authenticate the algorithm's retrieval effect in the image database.3. At present, the great development of science and technology has brought the information age. The sensor can obtain and transmit information effectively, but as the sensor number and variety has increased dramatically, users find it is becoming more and more difficult to select satisfied sensors from the sensor image library, so establishing an effective and perfect sensor image retrieval system is particularly important. A sensor image retrieval system based on the Matlab software is built by this paper. It realizes effective integration of system modules and a simple design of the user interface, and builds a sensor image database. The proposed image retrieval algorithm is applied to the sensor image library, combining with color features. The analytic hierarchy process (ahp) is used to set the different weight ratio in retrieval to verify the performance of the sensor image retrieval system. It finally proves that the system has rotation invariant features of the characteristics and better average precision in the sensor image library.
Keywords/Search Tags:Curvelet transform, GLCM, LBP, Analytic Hierarchy Process (AHP), Sensor image retrieval system
PDF Full Text Request
Related items