As introduced in Chapters 1 and 2, visual attention is one of the most important mechanisms of the HVS. However, other characteristics of the HVS have also been revealed by physiologists and psychologists. One such characteristic is visual masking, and it is reflected by the concept of just noticeable difference (JND) in image processing. The JND captures the visibility threshold (due to masking) below which the change cannot be detected by the majority of viewers (e.g., 75% of them).
Visual attention and JND are different: an object with high attention value does not necessarily mean that it should have a high or low JND value, since either case is possible. If the object is attended, the visibility threshold is lower than that for the case when the object is out of focus. Visual attention and JND are both related to the overall visibility threshold. Therefore, visual attention and JND are considered simultaneously in many image processing applications to provide a more complete visibility model of the HVS, and we refer it as attention-modulated JND, or as the overall visibility threshold.
JND modelling is first briefly introduced, and then two techniques – non-linear mapping and foveation – are presented to combine visual attention with JND modelling.
Many factors affect the JND, to account for spatiotemporal contrast sensitivity function (CSF) , contrast masking , pattern masking (including orientation ...