Content-Based Image Retrieval Using Wavelet

CBIR(Content-Based Image Retrieval) which becoming more and more important technique in large image database organization and management is the combine of the image retrieval and traditional multimedia information retrieval. It approximately can be divided into many aspect, shape-based retrieval, color-based retrieval, texture-based retrieval, and so on.Wavelet analysis is the widely used time-frequency tool. The wavelet coefficient which produced by wavelet transform can best display part character of the image. This article researched the usage of the Wavelet Analysis in computer image processing area, Image -Resuming, Image-Enhancing, Image-Dividing, Image-matching, Image-reconstruction, Image Retrieval, etc. Especially, this article pay more attention to application of the wavelet analysis used in the Shape-Based Image Retrieval.Shape is the important feature of object in image and shaped-based image retrieval can improve the accuracy and efficiency of image retrieval. In the Shape-based Image Retrieval, transforming the image with wavelet modulus maximum to get multi-scale edge images, then employs a set of seven invariant moments to extract the features of image, which can effectively resolve the variance with respect to translation, scaling, and rotation. Although the image's modulus maximums have been greatly compressed, but they still contain the redundant data, which is the information used in lower-scale modulus images has been used in higher-scale modulus images. To this question, this article put forward a improving algorithm. The sticking point is when getting the wavelet modulus maximum, setting a limit value, only those larger than the limit value can be recorded, and the limit value is dynamic. So the remaining wavelet modulus maximum can best display the important characters of the image. After compute the seven invariant moments through using the wavelet modulus maximum of the image, Similarity is given by the Euclidean distance between two images' normalized moment vectors. At last, the algorithm is also tested with medical CT images in a database; the experimental results formulate the verity the effectiveness of the algorithm.