Projection Algorithm Based Image Decomposition Models in Besov Spaces

摘要

In this paper we present a new class of image decomposition models. Following Meyer's ideas in a total variation minimization framework, our model decomposes a given image into a sum u + u,whereuis a function of Besov spaces B_(p,p)~s(the structure component of /), and the noisy componentuis modeled by the homogeneous Besov spaceE' = B_(∞,∞)~(-1) The proposed models can be seen as generalizations of Aujol-Chambolle model. We discuss an absolute wavelet based treatment for the new modelswhich is a continuation of Dirk A. Lorenz work in image decomposition. Finally, we present numerical results on denoising of both real and remote sensing images.