基于残差分析的自适应曲线拟合滤波算法

摘要

Aiming at the accurate detection and removal of salt and pepper noise in single image,we present the residual analysis-based adaptive filtering algorithm using curve fitting.First,the algorithm detects the noise points in conjunction with the means of employing point neighbourhood as the differentiation region.Secondly,it adaptively selects the direction with minimum residual as the direction of full signal point,uses curve fitting to determine the grey value at the noise point.After that,it carries out multi-direction template expansion processing on the noise images with higher density,and achieves effective recovery of the contaminated data points.Experimental result shows that this method has stronger ability in detail protection and noise point removal.%针对单幅图像中椒盐噪声的准确检测与去除,提出基于残差分析的自适应曲线拟合滤波算法.首先,结合以点邻域为判别区域的方法检测出噪声点;其次,自适应地选取残差最小方向为全信号点方向,利用曲线拟合,确定噪声点处灰度值;再次,对密度较高的噪声图像,作多方向模板扩充处理,实现被污染的数据点的有效恢复.实验结果表明,这种方法具有较好的细节保护能力和较强的噪声点去除能力.