HEVC saliency map computation

Abstract

This paper investigates whether the information related to the human visual saliency is still preserved at the level of the HEVC compressed stream syntax elements. In this respect, a new saliency model, matched to the peculiarities of this emerging standard is defined. It consists of four elementary maps, describing the four main saliency features: intensity, color, orientation and motion. These maps are defined based on the energies of the luma and chroma coefficients, on the variations of the intra prediction modes and on the energy of motion vectors, respectively. They are fusioned according to 48 static and static-dynamic pooling formulas. The results are compared to three state-of-the-art uncompressed (pixel) domain as well as to the MPEG-4 AVC compressed domain saliency maps. It is brought to light that the HEVC saliency model outperforms (with singular exceptions) the state-of-the-art uncompressed domain and is as good as MPEG-4 AVC saliency model. We can thus state that, as its MPEG-4 AVC ancestor, although not designed based upon visual saliency principles, the HEVC compression standard preserves this human visual property at the level of its syntax elements.