ABSTRACT
This chapter discusses features in video compression standards that are or can be used to reduce the visibility of coding artifacts, such as blockiness, ringing and mosquito noise. It describes the principles behind human visual system (HVS) models and examines different approaches to using these models. The chapter explains several ways of incorporating perceptual-based metrics into the rate control mechanism. Most video compression systems are transform based, which means that image detail is transformed into a frequency or space-frequency representation that has desirable properties, such as energy compaction and better matching to HVS noise sensitivity. All lossy video coding systems add noise, or distortion, to the image, and the visibility of the noise is a key feature of the compression algorithm. Picture level control generally involves meeting a pre-assigned target bit rate for each picture type. Visual mask ability of the source frame can give a starting point for the quantization profile to be used for the frame.
