Research Of Texture-Based Image Retrieval Technology

As technologies of computer and web are widely applied, the fields ofpublishing industry, medical and health work, architectural and engineeringdesign produce lots of digital images every day. How to analyze, store andretrieve these images is an imperative problem. The content-based imageretrieval technology can effectively resolve the problem and it also becomes ahotspot in study. The texture is a primary feature of image and an importantmethod in the content-based image retrieval system, so the paper mainly focuseson the texture-based image retrieval techniques.Firstly, the algorithms of texture-based image retrieval are researched in thepaper. Texture-based image retrieval algorithms in spatial domain: perceptual-based Tamura texture feature is analyzed;Run-length texture analyzing algorit-hmm is improved Run-length accumulating algorithm. Methods in frequencydomain: by two groups of comparative experiments, it is proved that wedge andring feature of Fourier transform can better represent the texture feature of animage. Methods in space/frequency domain: tree-structured wavelet transform intexture-based image retrieval is selected to use by analyzing how the energyfeature changes of texture image with the wavelet analysis, this method canpromote the efficiency of retrieval.Secondly, the feature matching of image is explored helpfully. The paperdemonstrates the data models related to the similarity measurement methods andproposes two kinds of multi-feature combination structures. Furthermore, howto use relevance feedback in contend-based image retrieval system is discussed.Combining the specific research in the paper, some of distance measurementfunctions are selected.Finally, a texture-based image retrieval system which is expansible andexperimental is designed and realized. By using a sorting assessment method inthis system, the retrieval algorithms are evaluated.