Image Retrieval Based On Color And Shape

With the development of multimedia, internet and digital image technology, there are more and more resources of digital images available. How to rapidly and effectively search the desired images from large-scale image database becomes an active hot point in the area of retrieval research.In the dissertation, the exploratory research work has been done around the low-level feature extraction and relevance feedback technology.The main research work and innovation of this thesis are as follows:1. Several key techniques and algorithms of CBIR are deeply analyzed and discussed, such as, colors space, the low-level feature descriptions including color, shape, and texture, the similarity measured between the features and the evaluation methods of image retrieval algorithms.2. A novel image retrieval method based on color and shape is introduced in the paper. Firstly, the color of an image is quantified in HSV model and the quantization results are classified into different status. Then, the transition probability matrix is presented to describe the color change and color relative entropy is proposed as a color descriptor. Finally, the spatial distribution entropy of interest points is introduced to denote the shape feature. Experimental results show that the new method gives better performance than the other methods mentioned in the paper.3. In order to reduce the gap between the low-level features and image semantics, the relevance feedback is introduced in the image retrieval system. In the dissertation, the performances of SVMs with different kernel functions are tested and compared. Experimental results show that function of RBF-SVM has higher precision of retrieval than the others.4. A content-based image retrieval model is designed in the paper based on visual C plus and SQL Server 2000 database. The users can browse the database random or according to the image classes, retrieve images by an example, choose relevance feedback to enhance the retrieval performance, etc.