An Image Segmentation System Based on Thresholding.

摘要

The segmentation algorithm proposed in this paper is a complex form of thresholdsing which utilizes multiple thresholds. The algorithms consists of two major components: a threshold selection component and a relaxation component. The threshold selection component automatically selects a threshold so as to maximize the global average contrast of edges detected by the threshold across the image. This algorithm for threshold selection compares favorably with other methods for automatic threshold selection. The threshold selection algorithm can be applied recursively to select additional thresholds by ignoring any edges which have already been detected by previously selected thresholds. The relaxation is independently performed for each threshold selected. The resulting binary images are intersected to produce the final segmentation. This algorithm works well not only for simple images but also produces reasonable segmentations for complex images.