A Gray Level-Gradient-Based Two-Dimensional (2D) Otsu Thresholding Method for Image Segmentation

摘要

The traditional gray level two-dimensional (2D) histogram region-dependent method often produces over-segmentation and cannot meet the increasing demands of image segmentation. We propose a gray level gradient 2D histogram segmentation method. Furthermore, we put forward an improved 2D histogram Otsu thresholding method and a fast, recursive algorithm. The results show that the improved Otsu's method creates better segmentation results, extracts the object with a clearer boundary, obtains better homogeneity in the region, and has great resistance to noises. Compared to the traditional 2D Otsu method, the improved Otsu method has more advantages, including faster running speed and better segmentation results.