Wavelet transform-based texture segmentation using feature smoothing
Xiang-Fa Song, Zhi-Guo Chen, Chenglin Wen, Quanbo Ge · 2004
Textures are one of the basic features in visual searching and computational vision. In this article, most of the attention has been focused on feature improvement based on pyramid wavelet transform using feature smoothing, and the goal is to improve textured image segmentation results, especially along the borders of regions. An improved method to extracting texture features in the feature extraction stage is described. Texture features are first estimated based on pyramid wavelet transform coefficients. The estimated texture features are then smoothed by a quadrant filtering method to reducing the variability of the estimates while retaining the region border accuracy. We conclude via results and discussion on an international texture database.