Image Enhancement of Noisy Images by Using Fuzzy Data-Dependent Bilateral Filter

Removing noise while preserving and enhancing edges is one of the most fundamental operations of image/video processing. However, removing noise and edge enhancement are conflicting requests, thus it is difficult to realize these two requests at the same time. Zhang and Allebach have proposed the adaptive bilateral filter (ABF) in order to realize the nose removing and edge enhancement at the same time. Bilateral filter (BF) combines range and domain filters based on Gaussian kernels. In ABF, range filter is changed depend on the output of Laplacian of Gaussian (LoG) operation. Since, LoG operation can detect edges from noisy images; ABF can remove noise while enhancing edges. However, in low signal to noise ratio (SNR) condition, LoG operation does not work well. Thus, the ability of ABF is decrease in low SNR condition. This paper presents the data-dependent BF based on fuzzy inference. We introduce the shape information based on local statistics and the estimated noise amplitude to fuzzy inference in order to derive suitable range filter for each pixel. The experimental results show that the proposed approach can effectively reducing the noise while enhancing edges under a wide range of circumstances.