摘要
考虑到图像存在异常像素,将邻域因素考虑在内对模糊C-均值聚类算法(FCM)和异常像素检测算法(APD)提出了改进。首先,提出了邻域因素的邻域-FCM(N-FCM),然后,提出了模糊异常像素检测算法(Fuzzy-APD)。实验过程中,选择噪声图像、彩色图像作为实验图像,对FCM和N-FCM算法进行性能比较,证实相比于FCM算法,N-FCM算法的收敛性明显提高,图像分割的正确率进一步改善;同时从图像中获取部分像素进行异常像素检测,实验证实相比于异常像素检测,Fuzzy-APD准确性更高。
Considering the existence of abnormal pixels in images,we take neighborhood factors into account,and propose the improved algorithms for both fuzzy C-means clustering(FCM)used in image segmentation and abnormal pixel detection(APD).Firstly,we proposed a neighborhood-fuzzy C-means clustering algorithm(N-FCM)about neighborhood factors,and then we proposed a fuzzy-APD algorithm for detecting abnormal pixels.During the experiment,noise images and color images were selected as experimental images.The performance comparison between FCM algorithm and N-FCM algorithm was carried out.It is confirmed that the convergence of N-FCM algorithm is significantly improved comparing with FCM algorithm,and accuracy of image segmentation is further improved.Meanwhile,some pixels are obtained from the image for abnormal pixel detection.Experiments show that the accuracy of the fuzzy abnormal pixel detection algorithm is higher than that of abnormal pixel detection.
作者
汪克峰
钱进
李仁璞
WANG Kefeng;QIAN Jin;LI Renpu(School of Computer Engineering,Jiangsu University of Technology,Changzhou 213001,Jiangsu,China)
出处
《实验室研究与探索》
CAS
北大核心
2020年第3期55-61,共7页
Research and Exploration In Laboratory
基金
江苏省自然科学基金项目(BK20161199)。
关键词
模糊C-均值聚类算法
异常像素检测
图像分割
邻域因素
fuzzy C-means clustering algorithm
abnormal pixel detection
image segmentation
neighborhood factors
