Robust feature extraction arithmetic based on parametrical log-polar transformation for texture image
Pan Hui · Journal of Zhejiang University(Science Edition) · 2009
This paper proposed a new type of feature extraction arithmetic for texture image, based on parametrical log-polar translation. This arithmetic is robust to geometric distortions. During the feature extraction, the autocorrelation image was introduced to eliminate the effect of translation, while the improved log-polar was used to eliminate the effects of rotation and scaling. In the retrieval process, a set of incomplete tree structured wavelet transform of the log-polar transformed autocorrelation image was used to measure similarity of texture images. The experiment results showed that the average retrieval rates of this arithmetic is about 81.05%, which is robust to the geometric distortions as compared with the typical algorithms.