Research On Content-Based Image Retrieval Technology

Content-based multimedia information retrieval technology has been widely applied in the cognitive science,artificial intelligence,pattern recognition,image processing, information retrieval,and other fields.Currently many achievements have obtained in the science research and commercial applications.In recent years,content-based image retrieval technology become one of the foreland research topic about computers and related disciplines. We should input the annotated image information before search in the method of traditional image retrieval.The deficiency of artificial annotating is that it needs abundant manpower, especially now a great deal of image information has produced everyday.If we always follow the old way we must not meet the need.Another deficiency is that artificial annotating can not express the contents which are very difficult to use word to describe clearly in image data.The purpose of this research is that building the image processing system of the intelligent and efficient large-capacity data retrieval by researching the model of user's interests with content-based image retrieval technology,thereby improving the accuracy of the search,has broad application prospects.Image Retrieval system performance greatly depends on the underlying visual image feature extraction and description.Image color,texture and shape,and other characteristics of the bottom are extracted and integrated in this paper,then a new space-based HSV 26 non-uniform color quantization algorithm is put forward and it accords with the subjective visual model.The algorithm compares with the traditional color quantization algorithm,we can draw an inclusion that it reduces the dimension of feature vectors,computation,less affected by the intensity of illumination,and improve the accuracy of the retrieval.There is Semantic gap between the physical characteristics of the lower image and top-level semantic features.Image semantic automatically decreases the accuracy of tagging with the increase of semantic image with the increase of semantic image.Image Semantics Automatic annotates by building the model of user's interests and integrating more than one visual features.Image semantic annotation is the same as optimization problem,a new method has been taken out which achieved the function.Finally,it summarizes the full text and pointes out the orientation of development in the future.