Fuzzy logic approach for removing noise

To avoid distorting the visual information in an image, smoothing algorithms must preserve the details. However, some fine details are frequently treated as noise by general smoothing filters. In such situations,we encounter the problem of ambiguity or uncertainty between noise and fine details. A new nonlinear smoothing method using fuzzy reasoning techniques is proposed. Fuzzy set theory [1,2] is a mathematical tool in modeling ambiguity or uncertainty and has been applied in image smoothing [3,41 and texture analysis [5]. In this paper, we use the definition of uniform surface uncertainty [3] as the fuzzification function. The uniform surface uncertainty, which ranges from 0 to 1, for a point p in the image is considered as the degree of belongingness of p to a uniform physical surface. Therefore, we can transform a grayscale image into a fuzzy image by using the fuzzification function. Then, we input the fuzzy image to fuzzy-rule-based system and produce a new fuzzy image. The processed fuzzy image is transformed back into a new grayscale smoothing image by the inverse of the fuzzification operator.