A DCT-domain JND model based on visual attention for image

Abstract

Most of the traditional JND models in DCT domain compute the JND threshold by incorporating the spatial contrast sensitivity function, the luminance adaptation effect and the contrast masking effect. How to integrate visual attention effect into the traditional JND models is still an open problem. In this paper, we proposed a new DCT-domain JND profile, in which a combined modulation function is built, based on the image saliency and textural characteristic to describe the visual attention effect and contrast masking effect on JND Threshold in DCT domain. Experimental results show that the proposed model can tolerate more distortion with the same perceptual quality, compared with the latest DCT-domain JND model. In terms of PSNR, the improvement of tolerated distortion is 0.54dB on average.