Designand Research Of Mobileimage Tetrieval

With the development of technology and information, more and more people are not satisfied with the retrieval which rely only on text to search for information, more and more attention has been paid on the content based image retrieval. This paper is devoted to the study and the completion of the content based image retrieval model. The model is composed of two parts:client and server. This paper mainly introduces the realization process of image retrieval system. The system consists of5modules:search the image input, image preprocessing, image feature value extraction, feature matching calculation and search results show. This paper puts forward the following four points of innovation in this process:1.In the design of the model, firstly,image preprocessing is used.After the image segmentation, the speed of Feature extraction is increasing. Finally, the matching precision and time of the whole system also had the very big promotion.2.The paper starts with Pyramid Smoothing Algorithm and acceleration scenarios that decor relate smoothing radius and calculation time to maintain relatively high real-time performance with sufficient smoothing effect. The overall performance make it preferable for image pre-processing as an alternative to conventional noise reduction algorithm and image filtering as an option, especially for cases with limited computational resources and large smooth radius.3.In the server’s feature extraction, feature extraction using CEDD and SIFT combination algorithm, two algorithms complement each other, improve the feature extraction accuracy, and finally improve the system accuracy. The CEDD and SIFT were reduced rank processing, firstly CEDD is from144dimensions to60dimensions, and SIFT extreme point information is from128dimensions to20dimensions. Finally, in order to integrate two algorithms, The BOW model used in the dimensionality of the SIFT, The feature vector is fixed, and the actual match method is relatively simple.4.In the process of implementation of multi-layer retrieval model, the first regional color histogram algorithm is used, in order to reduce the need for fast retrieval, next, we use improved CEDD features extraction in order to stream line the scope. Finally, the model use the algorithm combined with improved CEDD and SIF make the final search precision matching.In this paper, We have basically achieved the design of the mobile image retrieval model. We analyze the performance of the algorithm In two aspects of time and accuracy. The results show that the improved algorithm is successful and the multilayer retrieval system achieves the desired effect.