Level set segmentation depending on adaptive local information

A robust and efficient level set segmentation method using the energy in a reasonable range of segmentation dependent information (RSDI) is presented. A feature describing edges´ illegibility, called the edges´ blur descriptor, is first defined. Then it is used to compute RSDI adaptively and approximately. Some methods of segmentation dependent information localization (SDIL) utilizing a window function are discussed. The window size is determined by RSDI. We used the proposed method to improve the region-scalable fitting (RSF) method. Experiments show that it can choose RSDI effectively, and the segmentation accuracy and performance are superior to RSF method.