Abstract: This paper presents a new adaptive algorithm.The algorithm,which is based on static theory and fuzzy mathematics,evaluates the average and variance of each local data group(or named filter window),then uses the results to automatically select thresholds for each window to detect and smooth impulse noises.The experimental results show clearly that the growth of SNR using this algorithm is much higher than those of other common algorithms and similar algorithms whenever the density of impulse noise is below 5%,and the algorithm is more suitable for processing steadily changed signals.Compared with the traditional algorithms,our algorithm can automatically select thresholds,protect signals and produce higher SNR.The experimental results show that a higher SNR can be obtained by using this algorithm with the impulse noise density below 5%.