为了克服Otsu算法对图像分割的不足,在二维Otsu的基础上提出了基于二维直方图双斜率划分的快速Otsu分割算法。该算法将二维直方图上像素点之间的关系转换为斜率值,构建斜率值与频率的一维直方图,利用双阈值Otsu对灰度斜率的一维直方图取双阈值,利用求出的双阈值(临界斜率)对图像二维直方图进行重新划分;对划分的图像二维直方图的有效区域,再次利用一维Otsu方法划分,从而得到最终结果。最后通过实验验证此方法有较强的抗噪性,比二维Otsu方法计算效率高,处理效果好。
In order to overcome the disadvantage of Otsu algorithm on image segmentation, this paper proposed a fast Otsu algorithm based on the dual-slope division of two-dimensional histogram. The algorithm transformed the relations of pixel points in a two-dimension histogram into slope values and built a one-dimensional histogram of slope values and frequencies. By the use of a dualhreshold Otsu algorithm, the double threshold of the one-dimensional histogram of gray and slope values could be gotten. Then the two-dimensional histogram of image could be redistricted by the double threshold (critical slope). The final results could be gotten by using the one-dimensional Otsu algorithm again to divide the effective area of the redistricted two-dimensional histogram. The experiment results show that the proposed algorithm can achieve better anti-noise performance, computing efficiency, and processing effect than two-dimensional Otsu algorithm.