Fuzzy density weighted support vector regression for scene denoising

Support Vector Regression (SVR) is effective for image denoising. However, samples contaminated by noises are regarded as sparse samples which is poor fitted, and SVR can´t work well in case of noise level is too far from it in the training phase or noise density is high. In this paper, we propose a new fuzzy density weighted support vector regression (FDW-SVR) to address the problem of uncertainty of sample density and neglect of local similarity in images. The proposed FDW-SVR using a novel strategy of to design the learning weights, which is similar to the selection of human on the sample density. To handle the uncertainty of sample density, the learning weights (FDW) in the FDW-SVR are deduced using an interval type-2 fuzzy logic system (IT2FLS), which is an extension of the previous weighted SVR. Extensive experimental results demonstrate that our method can obtain better performances in terms of both subjective and objective evaluations than those state-of-the-art denoising techniques.