近似测地距离度量下的图像抗噪分割方法

Noisy Image Segmentation Based on Approximate Geodesic Distance

对含有噪声的图像进行有效分割是图像处理中的难点问题之一.为解决欧氏距离带来的空间局限性,对含有噪声的图像进行有效分割,提出一种基于近似测地距离和边界加权的重心Voronoi图划分(CVT)能量模型的图像抗噪分割方法.首先利用图像梯度的大小和方向建立一种近似测地距离计算模型,降低了算法的时间复杂度;然后采用该测地距离测度定义边界加权的CVT能量模型,并极小化能量模型实现数字图像的抗噪分割.实验结果表明,该方法可以有效地对含有噪声的数字图像进行抗噪分割.

Segmentation for noisy images is a difficult topic in image processing. To break through the restriction of Euclidean distance and segment the noisy image effectively, a new method based on a geodesic framework and EWCVT （edge-weighted centroidal Voronoi tessellation） energy model is presented in this paper. Firstly, we propose an approximate model of geodesic distance according to image gradient, which can decrease the computation complexity of the algorithm greatly. Then, we apply this geodesic distance to achieve anti noisy image segmentation by minimizing EWCVT energy. Experimental results show that the proposed method can carry out anti noisy segmentation effectively.