关键词: 隶属度场, Gibbs随机场, 邻域标准差, 自适应加权

Abstract: Membership field of traditional fuzzy C means algorithm consider gray information of single pixel only，which is beneficial to retain details but weak in denoising an image.On the other hands，Gibbs random field depicts the spatial distribution of pixels，which is beneficial to smooth noise but poor at retaining image details.Thus，an improved method is proposed to take advantages over the two algorithms respectively，which can automatically determine the possible location of a pixel and adjust the proportion of the two fields according to neighborhood information of a pixel.Experiments show that the improved algorithm can adjust the weight of two fields adaptively to remove the noise while preserving more details.

Key words: membership field, Gibbs random field, standard deviation of neighborhood information, adaptive weighting