摘要
传统二维Otsu阈值分割算法未考虑人类视觉特性,分割结果不符合人眼视觉感受。为此,提出一种二维Otsu算法与侧抑制网络相结合的分割算法。该算法从基于人类视觉系统的侧抑制网络出发,利用侧抑制网络增强中心,抑制周围的特性,通过侧抑制网络处理原始图像,得到侧抑制图像,构建基于像素的灰度信息和侧抑制信息的二维直方图,并采用类间最大方差作为最佳阈值的选取准则。实验结果表明,与传统的Otsu算法和二维Otsu算法等相比,该算法具有较好的对比度、光照强度适应性和间断拟合能力,并能提高对图像噪声的鲁棒性,获得更理想的分割结果。
The traditional two-dimensional Otsu thresholding segmentation algorithms do not think about human vision characteristics and the result of segmentation can not match up to the visual perception of human eye. In order to solve this problem,an algorithm based on the two-dimensional Otsu algorithm and the lateral inhibition network is proposed. In this algorithm, the lateral inhibition network of human visual system that has the features of enhancing center and inhibiting surroundings is fully used. The lateral inhibition network is utilized to process the original picture and obtains the lateral inhibition picture. A two-dimensional histogram based on the gray information and lateral inhibition information of pixels is established. The maximum between-cluster variance is chosen as the criterion to select the optimal threshold. Experimental results show that this algorithm not only is well adapted to the contrast and illumination intensity, but also has the capacity for fitting the breaks compared with the traditional Otsu algorithm and two-dimensional Otsu algorithm. It improves the robustness to image noise and obtains more perfect segmentation results.
出处
《计算机工程》
CAS
CSCD
北大核心
2015年第6期195-200,共6页
Computer Engineering
基金
国家自然科学基金资助项目(60975021)
湖北省自然科学基金资助项目(2012FFB07404)
武汉大学软件工程国家重点实验室开放课题基金资助项目(SKLSE20120934)
中南民族大学中央高校基本科研业务费专项基金资助项目(CZY12007
CTZ12023)
关键词
侧抑制网络
二维直方图
OTSU算法
阈值选取
阈值分割
lateral inhibition network
two-dimensional histogram
Otsu algorithm
threshold selection
threshold segmentation
