首页|自适应空间强度约束和KL信息的模糊C均值彩色噪声图像分割

自适应空间强度约束和KL信息的模糊C均值彩色噪声图像分割

扫码查看
原文链接
  • NETL
  • NSTL
  • 万方数据
为了增强传统模糊C均值聚类算法的抗噪性能，保持任意像素与相邻像素间的隶属度相似性，提出一种自适应空间强度约束和KL信息的模糊C均值彩色噪声图像分割算法。首先，通过快速双边滤波器获取局部空间强度信息，用于平滑噪声像素;然后，将局部加权平均隶属度作为先验概率，并通过KL信息将其嵌入目标函数，从而优化隶属度的划分矩阵;最后，计算原始图像与双边滤波图像间的绝对强度差，用指数形式的绝对强度差作为双边滤波图像的自适应权值，并将其倒数作为原始图像的自适应权值。当混合噪声密度为30％时，所提出算法在彩色合成图像上的划分系数和划分熵分别为99。66％和0。58％，在彩色真实图像上的划分系数和划分熵分别为98。77％和2。03％。实验结果表明，与其他相关算法相比，所提出算法的抗噪性能更强、分割精度更高、稳定性更好。
Fuzzy C-means with adaptive spatial intensity constraints and KL information for color noise image segmentation
In order to enhance the anti-noise performance of traditional fuzzy C-means clustering algorithms and preserve the membership similarity between arbitrary pixels and neighboring pixels,a fuzzy C-means color noise image segmentation algorithm with adaptive spatial intensity constraints and Kullback-Leibler(KL)information is proposed.Firstly,the local spatial intensity information is obtained using a fast bilateral filter,which is used to smooth the noisy pixels.Secondly,local weighted average membership is taken as the prior probability,and it is embedded into the objective function by KL information,so as to optimize the membership partition matrix.Finally,the absolute intensity difference between original image and the bilateral filtered image is calculated,and the absolute intensity difference in exponential form is adopted as adaptive weight of the bilateral filtered image,and then its inverse is applied as adaptive weight of original image.When the mixed noise density is 30％,the partition coefficient and partition entropy of the proposed algorithm are 99.66％and 0.58％on the noise synthetic image,and then 98.77％and 2.03％on the real noise image,respectively.Experimental results show that the proposed algorithm has stronger anti-noise performance,higher segmentation accuracy and better stability in comparison with other related algorithms.

image segmentationfuzzy C-meansKL informationspatial intensity informationthe mixed noise

彭家磊、黄成泉、雷欢、覃小素、陈阳、周丽华

展开 >

贵州民族大学贵州省模式识别与智能系统重点实验室,贵阳 550025

贵州民族大学数据科学与信息工程学院,贵阳 550025

贵州民族大学工程技术人才实践训练中心,贵阳 550025

图像分割 模糊C均值 KL信息 空间强度信息 混合噪声

国家自然科学基金项目贵州省省级科技计划项目贵州省教育厅自然科学项目贵州省模式识别与智能系统重点实验室2022年度开放课题项目

62062024黔科合基础-ZK[2021]一般342黔教技[2022]015GZMUKL[2022]KF03

2024

10.13195/j.kzyjc.2023.1032

控制与决策
东北大学

控制与决策

CSTPCD北大核心
影响因子：1.227
ISSN：1001-0920
年,卷(期)：2024.39(10)