Using Wavelet Transformation in Information Retrieval System

While many content-based retrieval systems are currently available, different searching algorithms were being applied to increase the accuracy of the result. However, there are still some key issues yet to be resolved; the effectiveness and accuracy of these existing systems are still in doubt, especially when dealing with video contents. Since video frames contain similar color definitions from one to another, color-based retrieval algorithm may fail to regconise different video clips. In this book, the Wavelet based Retrieval was reviewed, including Haar wavelet and Daubechies D4 wavelet. Various different types of clustering techniques were also applied to the retrieval, such technique includes One-level clustering using Forgy, K-Means and Isodata, as well as Two-level clustering with 2, 3 and 5 partitioning. The experiments were conducted by a number of evaluations over the image database, which contains more than 2,500 images with the mixture of both still images and video sequence images.