Segmentation

The aim of this course is to present the main families of approaches to image segmentation, i.e. the partitioning of an image into different zones, each representing a characteristic object. The course is divided into three parts: general information on segmentation methods, segmentation methods based on the detection of homogeneous regions, and finally, approaches based on the detection of heterogeneities in the image, i.e. on the detection of boundaries.

TS206

1/ General Definitions Dealing with non-stationarity: segmentation Segmentation's place in a signal or image processing chain 2/ Region segmentation Thresholding Segmentation by Division/Merge Segmentation by Markov model. Segmentation by growth around a seed Segmentation by Watershed 3/ Contour segmentation General information on derivative methods Sobel, Prewitt and Kirch masks MDIF and NAGDIF operators Laplacian and other second-order filters An optimal FII contour extractor: Canny's filter Canny's extension to the FIR case: Deriche's operator

Image processing