结合边缘信息的二维Otsu阈值分割算法研究

Two-Dimensional Otsu Threshold Segmentation Algorithm Based On Edge Information

针对传统二维直方图灰度———平均灰度法存在的区域错分,导致图像分割质量和抗噪性能的不足,提出了一种结合边缘信息的二维直方图灰度-非局部均值的算法。非局部均值能综合利用各像素点与其邻域点、邻域块之间的相关信息,降低图像受噪声干扰的影响。在二维直方图区域划分中,该算法通过边缘检测算子提取出图像边缘点,利用一定约束下的两个边缘点与主对角线,来确定噪声和目标背景区域的界线。通过状态转移算法结合截距阈值法求取阈值,实现了降维简化处理。实验表明,该算法能获得更好的图像分割效果。

In view of the region misclassification of traditional gray-the average gray level 2D method,which leads to the shortage of image segmentation quality and anti-noise capacity,A gray-non-local means level 2D method based on edge information is proposed.Non-local means can make use of the relevant information between each pixel and its neighborhood points and neighborhood blocks to reduce the impact of noise.In the area division of 2D histogram,the algorithm extracts the edge points of the image by the edge detection operator,and uses the two constrained edge points and the main diagonal to determine the boundary line of noise and target background area.The threshold is solved by the combination of state transition algorithm and intercept threshold method,the dimension reduction and simplified processing are achieved.Experiments show that this algorithm can achieve better image segmentation result.