Circle-Valued Nonlinear Diffusions with Application to Texture Segmentation
Xiaogang Dong, Ilya Pollak · 2006
We introduce a novel nonlinear diffusion equation for images defined on an arbitrary graph and taking values on a circle. We use this nonlinear diffusion to develop an effective segmentation algorithm for such images. We apply this algorithm to segmenting texture images based on the orientation feature defined as the angle of the gradient.