摘要
脑核磁共振图像常常受到噪声的影响,且有灰度不均、边界模糊的特点,使得传统聚类算法无法获得理想的脑部肿瘤分割结果,为此提出一种基于边缘保持滤波和改进核模糊聚类的脑肿瘤图像分割方法.该方法首先采用改进的引导滤波算法对图像进行预处理,解决平滑图像时不能保留图像边缘的问题;然后将传统核模糊C-均值聚类算法(Kernel fuzzy C-means clustering)中的单一高斯核函数替换为混合高斯核函数,将数据由低维空间映射到高维特征空间;最后将马尔科夫随机场的先验概率引入,对算法的目标函数进行修正,进一步增强算法的抗噪性.实验结果表明,所提方法在去除噪声的同时,能够有效保留图像的边缘信息,PSNR值相比传统算法提升0.8041~2.0962 dB,SSIM值相比传统算法提升0.0312~0.0654,且算法分割精度更高,Dice指标和Jaccard指标的平均值分别达到0.9551和0.9141.
Brain MRI images are often affected by noise,and the gray level is not uniform and the boundary is fuzzy,which make the traditional clustering algorithm can not obtain the ideal results of brain tumor segmentation.Therefore,a brain tumor image segmentation method based on edge preserving filter and improved kernel fuzzy clustering is proposed.In this method,an improved guided filtering algorithm is used to preprocess the image to solve the problem that the edge of the image cannot be preserved when the image is smoothed.Then,the single Gaussian kernel function in the traditional kernel fuzzy C-means clustering algorithm is replaced by the mixed Gaussian kernel function,and the data is mapped from the low-dimensional space to the high-dimensional feature space.Finally,the prior probability of Markov random field is introduced to modify the objective function of the algorithm to further enhance the anti-noise performance of the algorithm.Experimental results show that the proposed method can effectively retain image edge information while removing noise.Compared with the traditional algorithm,the PSNR value is improved by 0.8041~2.0962 dB,and the SSIM value is improved by 0.0312~0.0654.Moreover,the segmentation accuracy of the proposed method is higher,and the average values of DICE index and Jaccard index reach 0.9551 and 0.9141,respectively.
作者
王志刚
冯云超
WANG Zhi-gang;FENG Yun-chao(College of Computer Science and Engineering, Hunan Normal University, Changsha 410081,China)
出处
《湘潭大学学报（自然科学版）》
CAS
2021年第3期114-126,共13页
Journal of Xiangtan University(Natural Science Edition)
