Fourth order method for image denoising based on Gaussian curvature

摘要

A novel fourth order PDE model for image denoising is presented in this paper. Gaussian curvature is integrated into the construction of geometrical energy functional within within the geometrical framework for image processing. The corresponding time dependent PDE model is a combination of three fourth order terms and a third order term with some feasible constrains. Image denoising is attained by surface evolution driven with Gaussian curvature. The new model avoids staircase effect and preserves edges while smoothing noise. Experimental results are given to demonstrate the performance of the proposed method.