Impact of Redundancy and Gaussian Filtering on Contourlet-Based Texture Retrieval

摘要

Multiscale image representations, such as contourlets and wavelets are crucial to significant feature extraction for texture retrieval. In this paper, the aim is to highlight the positive impact of added redundancy and Gaussian filtering in multiscale image decompositions for texture retrieval. Using energy-based retrieval framework on Vistex database, conducted experiments on five multiscale transforms (contourlets, wavelets and their redundant counterparts), show the competitive enhancement provided by the redundant contourlet decomposition in terms of discriminant features and improvement of retrieval rates.