Scale-adaptive orientation estimation

This paper focuses on directional textures. It provides a new framework for the design of convolution masks dedicated to orientation estimation and an adaptive algorithm which chooses the best mask size for each pixel. The design of the adaptive algorithm is based on the combination of two complementary operators: a gradient based operator which is adapted to sloped regions, and a ´valleyness´ detector which fits the crests and valleys. Each operator is optimized in terms of bias reduction. The scale adaptive implementation of the operator is carried out in two steps. First, characteristic points are detected. Their relative positions provide us with an estimation of the scale. Next, the size of the convolution masks is chosen according to the estimated scale. Experiments on synthetic and natural textures are provided and show the efficiency and relevance of our approach.