一种基于鲁棒估计的自适应图像平滑方法

基于鲁棒统计理论,提出一种新的保持图像边缘的图像平滑算法.将图像滤波与平滑转化为一个分段常数模型的鲁棒估计问题.并由此导出基于分段常数模型的尺度自适应鲁棒滤波器.为了自动确定鲁棒滤波器的最优尺度参数,使用直方图分析,结合非参数统计的方法来确定各点的鲁棒尺度参数.该算法可以在平滑图像的同时,保持原始图像中的边缘结构.实验表明,该算法平滑图像的效果,明显优于固定尺度的平滑算法.

Adaptive image smoothing based on robust estimation

Based on robust statistics,a novel edge preserving image smoothing algorithm was proposed. The problem of image filtering and smoothing was solved by robust estimation for piece-wise constant model, from which the scale adaptive robust filter was derived. To obtain the optimal scales of the robust filter, histogram analysis was adopted with nonparametric methods. This smoothing algorithm can preserve the structure information in the original image while smoothing it simultaneously. Experiments show that the proposed algorithm outperforms the fixed scale smoothing algorithms.