ABSTRACT
This chapter presents the computational just-noticeable difference (JND) models in both subbands and pixels, under an integrated formulation. It provides a systematic introduction in the field to date, as well as a practical user’s guide for the related techniques. JND has been used to determine not only the noticeable visual distortion but also the possibly noticeable visual quality enhancement. Two major factors have been considered for the spatial JND in pixel domain: luminance adaptation and texture masking. The chapter introduces the models with discrete cosine transform subbands with a general and easily-adopted formulation. In various image processing tasks, pixels are processed for visual quality improvement, compact signal representation or efficient data protection. Digital images are acquired, synthesized, enhanced, watermarked, compressed, transmitted, stored, reconstructed, evaluated, authenticated, displayed, or printed before being presented to the human visual system. In psychophysical studies, the relationship between an external physical stimulus and its psychological representation in the mind has been investigated.
