Abstract: An image adaptive K-nearest neighbour filtering algorithm（AKNNMF） is presented by the exploiting of luminance，edge，texture characteristics of the host image.Firs possible noisy pixels are determined according to the just noticeable difference based on Human Visual System（HVS） masking characteristics，and then the filter window size and the K value is adaptively adjusted by the noise density.Finally，the noisy pixels are removed by the improved adaptive K-nearest neighbour filtering algorithm.It can successfully remove salt and pepper noise as well as preserving image detail.Experimental results show that the presented scheme is superior to standard median filters，mean filters，K-nearest neighbour mean filters.