Research On Image Retrieval Based On Its Texture

With the development of multimedia, the technology of image retrieval becomes a research hot in image processing. Content-Based Image Retrieval (CBIR) is the one who matches the similarity by the features of image content itself. Nowadays, the research work focus on the features extracted from the color, the texture or the shape of image. In this thesis, a new algorithm based on the adaptive lifting scheme is proposed for texture image retrieval. It can solve the problems in the algorithm of image retrieval based on wavelet domain.Much noise would be generated during the image obtaining and transformation. In the image decomposing, there would be a lot of microscopic abrupt changes produced by noise pollution. It will influence final retrieval results because of the improper decomposition by traditional wavelet. A new descriptor named Gradient Amplitude Descriptor (GAD), which aims at improving the adaptive wavelet lifting scheme based on Neville filters, is proposed for the image adaptive decomposition in this thesis. The final retrieval result would not be affected by matching the features extracted from the decomposing result what is similar to the result of clean image can be obtained by this descriptor.Even obtaining the ideal decomposing result, it is very important to extract feature in image retrieval. The extracted feature should not only preserve geometrical transforms invariance, but also visual invariance. In this thesis, a new feature named synthetic Sorted Gradient Direction Histogram (SSGH) is designed by gradient calculation of wavelet coefficients. Comparisons with prior methods show superiority in achieving both robustness to geometrical variations and optimum acutance matching to human visual perception.