Improved adaptive median filtering
一种改进的自适应中值滤波方法

An improved adaptive median filter is proposed to restore images corrupted by impulse noise. There are mainly three improved points in this algorithm. First, the image pixels are classified into signal pixels and noise pixels according to the decision criterion. Second, a measure, denoted as minimum set distance, is introduced to avoid misclassification of high frequency signal as noise. Third, the median of minimum uncorrupted pixel set is used to restore noise pixel, in order to eliminate its neighborhood noise impact. The results of comparison experiments with RAMF and NASMF demonstrate that the proposed method can detect noise with high accuracy, remove noise efficiently while retaining image details, especially to images with high noise density.