Research On Retrieval And Recognition Algorithm For Color Image

Along with the rapid development of Internet, data mining, Image processing and computer vision technology, more and more attention has been paid to image recognition and retrieval and classification technology, and there has been a great development of image retrieval technology. The color image data is the world’s most widely used image data of all the image data. In recent years, how users can quickly and accurately find what they want from the color image resources and a large number of images have become an important research topic and a variety of image retrieval and recognition algorithms emerge endlessly. Among all of them, the content based image retrieval(CBIR) is a hot research direction in image retrieval.This paper is mainly about the research of the content based color image retrieval and recognition. Content based color image retrieval is a retrieval technology based on the content, namely semantics and characteristics(such as: color, texture, shape, position and so on) of the object in color image. In recent years, the content based color image is one of the most active research directions in the field of computer vision, digital image processing and data mining. At the same time, because of the fast development of the technology of data mining, a large number of excellent clustering and classification methods have been introduced to image retrieval and recognition process. This paper is focused on this aspect of the research in image retrieval and recognition problem with the method of clustering and classification, and the main research results are as follows:1. Affinity propagation(AP) clustering simultaneously considers all data points as potential exemplars.It takes similarity between pairs of data points as input measures,and clusters gradually during the message—passing procedure. But the result of AP clustering Algorithm in the data set of complex structure(non-group)is not very good. Therefore, we proposed a new clustering Algorithm by adding a merge process on the basis of AP clustering Algorithm, called M-AP algorithm which can effectively solve this kind of problem. When the number of samples is very large, we can effectively solve the problem of large sample by using CVM compression algorithm.2. Aiming at the problem of the content-based image retrieval, we propose a new image retrieval method based on AP clustering. By applying the AP clustering algorithm into the image of the same class, each class can draw several clustering center. Then we find the distance between the test images and each cluster center, the type of the test image is the same with the center which is nearest to the test image. Experiments of image retrieval system shows that: the image retrieval process based on the AP clustering algorithm is not only very efficient, but also can solve the retrieval problem of all kinds in a training time.3. Image retrieval system based on OBVM mainly focus on a technology called image feature classification included in image retrieval, systematically studying the image retrieval system based on classification of image features. The content is mainly covered by the classification of image features, such as SVM, OCVM, OBVM(Online ball Vector Machine). The HSV color features and hog features and PCA techniques has been used in order to run faster. And with the help of retrieval feedback learning mechanism, we can improve the discrimination accuracy of image in the image library. At the same time, we can improve the reliability of content-based image retrieval with the expanding of the scale. The experimental results show the effectiveness of the proposed method, the retrieval performance can be gradually achieved the best and has better practicability.