A Quantization Matrix Design using Total Variation for Sparse Representation-based Image Coding

This paper proposes a quantization matrix design method for sparse representation-based image coding. In the proposed method, we define the complexity of the atoms in the dictionary designed by K-SVD as a simple total variation, and design a quantization matrix to quantize the sparse coefficients based on the similarity with the complexity in the DCT atoms of HEVC. This method makes it possible to perform coefficient quantization adaptively for each atom's feature. Experimental results show that the proposed method provides higher coding efficiency in comparison with conventional uniform quantization matrix from a point of perceptual picture quality.