Fast multi-scale detail decomposition via accelerated iterative shrinkage

We present a fast solution for performing multi-scale detail decomposition. The proposed method is based on an accelerated iterative shrinkage algorithm, able to process high definition color images in real-time on modern GPUs. Our strategy to accelerate the smoothing process is based on the use of first order proximal operators. We use the approximation to both designing suitable shrinkage operators as well as deriving a proper warm-start solution. The method supports full color filtering and can be implemented efficiently and easily on both the CPU and the GPU. We demonstrate the performance of the proposed approach on fast multi-scale detail manipulation of low and high dynamic range images and show that we get good quality results with reduced processing time.