A Pseudo Cross Bilateral Filter for Image Denoising Based on Laplacian Pyramid

To attain the goal of preserving the edges and removing the noise for the images, a pseudo cross bilateral filter based on Laplacian pyramid is proposed. The proposed algorithm firstly decomposes the image and its filtered version into several levels via Laplacian pyramid, in which the filtered version is obtained by adaptive Wiener filter. At each level of the decomposed images, a pseudo cross bilateral filter is then applied regarding the filtered version as the "real" image. To optimize and determine the parameters for the algorithm, the noise variance estimation at each level and its relationship with the photometric spread parameter in bilateral filter are discussed. The experimental results indicate the proposed algorithm is superior to adaptive Wiener filter and the traditional bilateral filter in both objective and subjective assessment.