An integrated framework for both compression noise reduction and super-resolution

Even though super-resolution is a promising technique, it can cause unwanted increase of compression noises, if it is applied to an image coded with a low bit-rate. To solve this problem, we propose an integrated framework to reduce compression noises and to recover high-frequency components of a coded low-resolution image. The framework effectively combines two patch-based algorithms, a patch-matching-based 3-D filtering algorithm and an example-based super-resolution algorithm, based on the combination of 3-D transform coefficients. Experimental results demonstrate that the proposed framework successfully improves the resolution while alleviating compression noises in the images coded with low bit-rates.