噪声图像滤波的树型滤波器

Noisy Image Filtering Based on Tree-Type Filters

本文提出了一种在空间域根据图像的邻域特性和边缘方向信息进行滤波与增强的方法。该方法利用图像的结构信息,对噪声进行分区处理,克服了传统线性滤波器引起的边缘模糊和细节损失。在算法实现上,充分发挥中值滤波器滤除脉冲干扰能力强,边缘保持好的优点。从而使恢复图像边缘和细节清晰。理论分析和计算机仿真结果表明,本文提出的滤波器结构在噪声滤波和边缘保持方面优于中值滤波,α-裁剪均值滤波、Winsorized均值滤波,建立了一种崭新的硬件方案。(本算法已在PCVISION图像系统上仿真。)

This paper proposes a new filtering and enhancing scheme based on neighbour feature and edge direction in spatial domain. Structure information of 2-D image is used completely by dividing the whole image into some subimages, which overcomes edge blurred and details lost in a linear filter. On the algorithm, we use the median filter that srppresses impules noise and preserves edge well by dividing the filtering window, and consider the resolution of smooth region. Theoretical analysis and computer simulation results have shown that the performance of the filter proposed here is better than that of the median filter, α-trimmed mean filter and Winsorized mean filter in the noise suppressing and edge preserving. A hardware realization is discussed. This algorithm is simulated on the PCVISION image system.