Highly Nonlinear Approximations for Sparse Signal Representation
Enhancing sparse representation of color images by cross channel transformation
A method for enhancing sparsity in the approximation of color images by 2D atomic decomposition is proposed. It is demonstrated that, by applying a transformation across the direction of the RGB channels, the resulting 2D arrays are sparser in the wavelet domain than without the cross channel transformation. This feature is further exploited by approximating the transformed arrays using an effective greedy strategy with a separable highly redundant dictionary. The relevance of the achieved sparsity is illustrated by a simple encoding procedure. On a set of typical test images the compression at high quality recovery is shown to significantly improve upon JPEG and WebP formats. The results are at least as good as those produced by JPEG2000 and perceptibly superior for some of the images.
by Laura Rebollo-Neira and Aurelien Inacio
