Research On Bilateral Filtering Denoising Methods And Their Applications

Image denoising is a basic topic in the field of image processing and computer vision. The basic aim of image denoising is to remove noises while preserving image details as much as possible. However, suppressing noises and preserving image details is often a contradiction, which has not yet been well solved in image processing field. This dissertation profoundly analyzes some drawbacks of bilateral filter (BF) and proposes several improved BF algorithms. The proposed algorithms obtain better image denoising results than the traditional BF method and several existing BF methods. The main research contents in this dissertation are outlined as follows:1. The traditional bilateral filter method is analyzed intensively, and it is pointed out that it cannot preserve the edge feature of the images and it is not easy to be realized by hardware. Besides, its parameters need to be set based on experience. To tackle these problems, an improved adaptive bilateral filtering algorithm is put forward. A compensation function based on the similarity judgments is introduced in the algorithm to preserve the edge feature effectively, the Thiele’s continued fraction is used to approximate exponential function so that the algorithm can be better implemented in hardware and the parameters are chosen based on adaptive gray variance instead of experience. The experimental results show that the adaptive bilateral filter is superior to traditional bilateral filter or other modified bilateral filters in both subjective and objective evaluations.2. We also profoundly analyze some drawbacks of BF method in filtering parameter and propose an improved BF image denoising method. Based on singular value decomposition (SVD) in the gradient domain, local structure factor is constructed. Then the whole image is divided into flat edge areas and texture areas and different filter parameters are set for these areas respectively. Experimental results show that the proposed method outperforms original BF method and other existing BF methods based on preselection in terms of denoising results and running speed.