Small Probability Strategy Based Otsu Thresholding Method for Image Segmentation
-
摘要: Otsu自适应阈值法是一种经典的图像阈值分割方法,在其基础上发展起来的2维Otsu法及其改进算法由于存在计算(或空间)复杂度较高、抗噪能力差、难以扩展到多阈值等不足而制约了其应用。该文针对2维Otsu法的不足,将噪声点的出现视为小概率事件,用噪声点的邻域均值代替其灰度值,将噪声点转换为目标(或背景)像素,减少了图像中的噪声点数量;继而直接采用1维Otsu法进行分割,以较小的代价获得良好的分割效果。算法分析及测试实验表明:与现有2维Otsu法相比,该算法在复杂度、抗噪性、多阈值扩展性等方面都有明显改善。Abstract: Otsu adaptive threshold algorithm is a classic image segmentation method. The two-dimensional Otsu algorithm and its improvements which based on original Otsu algorithm are restricted, due to their computation(or space) complexity, inability for anti-noise, difficulty to extend to multilevel thresholding. In order to improve these shortages, regarding noise points appearances as small probability events, noise point is changed to objective(or background) pixel by using its neighborhood average gray level to instead its gray level. Then the processed image is segmented through one-dimensional Otsu. So this method obtain good performance at low cost. The experimental result shows that this method has significant improvements in complexity, ability for anti-noise, ability for extending to multilevel thresholding and so on.
点击查看大图
计量
- 文章访问数: 2284
- HTML全文浏览量: 132
- PDF下载量: 990
- 被引次数: 0