摘要 本文提出了一种新的基于极大似然法的椒盐噪声滤波算法。在传统BP 算法中引入

了极大似然估计，在训练样本时能够在考虑网络逼近行为的同时对噪声分布进行估计。而且

针对椒盐噪声模型构造了新的鲁棒误差函数，从而使算法本身的抗干扰性增强。实验结果表

明了该算法与传统BP 算法相比具有更好的滤波性能。-Abstract This paper presents a new maximum likelihood method based on impulse noise filter. In the traditional BP algorithm, the introduction of the maximum likelihood estimation, the training samples can also be considered acts of the network approach to estimate the noise distribution. Salt and pepper noise and model construction for a new robust error function, so that the enhanced anti-jamming algorithm itself. Experimental results show that the algorithm compared with the traditional BP algorithm has better filtering performance.