Segmentation of color noisy images using anisotropic diffusion

摘要

Various factors such as noise and miscellaneous small features that obstruct proper image segmentation generally exist in an image. We extend an anisotropic diffusion algorithm combining morphology to color images, that is an edge-preserving smoothing for grey scale images, as pre-processing of segmenting color noisy images. We improve a morphology filter by using new structuring elements and propose to select each element according to edge information. The proposed method shows the capability of providing good pre-processing for color noisy images. We also make a try to determine the diffusion coefficient and the number of diffusion automatically.