Research On The Key Technology Of Image Retrieval

This paper mainly discusses the development status of image retrieval, content-based image retrieval and the environment, as well as the major achievements in this field. Image retrieval algorithm based on texture appears in development process, color, gray scale, shape retrieval algorithms, but these algorithms of local feature description ability and fuzzy, to take into account the overall characteristics, in this process is based on DCT transform, wavelet transform, JPEG compressed domain image feature description, but because these algorithms limitations, by Sift algorithm is used as image feature description algorithm.Sift algorithm is used as image feature description algorithm, based on the application status, as the image retrieval algorithm key algorithm at the same time, the advantage and disadvantage to improve, to meet the requirements of image retrieval application scenarios, and combined with other algorithms finally get the content based image retrieval algorithm.In the image feature description of the process, the first use of Sift algorithm to extract image feature, through the filter is set to realize the image features of the initial filter. In the filtered image feature, application of K means algorithm to cluster features index, then using the Euclidean distance to limit excessive intensive feature points, which are descriptive of the image feature vector. For image feature description of high dimensional vector, using vector principal component analysis to establish efficient query, that the image retrieval process efficiency. The experimental results show the effectiveness of the algorithm.