Abstract: As to the color images corrupted by impulse noise，a series of iterative filtering algorithm based on noise detection is proposed.An impulse detector is presented to find out all pixels that are considered as noise before filtering.If a pixel is considered as noise，it will be replaced by a pixel in the neighborhood which is noise-free during filtering procedure later;otherwise，it is unchanged.The conventional filters are improved as the filtering algorithm later.The conventional filters include the vector median filter，the basic vector directional filter，and the directional distance filter.As demonstrated by the extensive experimental results，all the proposed filtering methods compare favorably with conventional techniques in the capabilities of noise attenuation and details preservation，in both quantitative and qualitative measures.

Key words: color image filtering, impulse noise, noise detection, iterative filtering algorithm

摘要： 对于被脉冲噪声污染的彩色图像，基于噪声检测，提出了一系列迭代滤波算法。运用脉冲噪声检测器，估计出图像中的噪声像素，应用一系列后续滤波算法，只对检测出来的噪声像素进行滤波，而对非噪声像素（即信号像素）保持其值不变。传统的矢量滤波算法（矢量中值滤波、基本矢量方向滤波和方向距离滤波）加以改进后可作为后续的滤波算法。实验结果表明，这些新的滤波算法与传统滤波算法相比，在有效消除噪声的同时，更能够保留图像中的边缘和细节特征。

关键词: 彩色图像滤波, 脉冲噪声, 噪声检测, 迭代滤波算法