ABSTRACT
This chapter introduces the background of the foveation feature of the human visual system that motivates the research effort of foveated image processing. It reviews various foveation techniques that have been used to construct image and video coding systems. The chapter provides a specific example of such systems, which deliver rate scalable codestreams ordered according to foveation-based perceptual importance, and has a wide range of potential applications, such as video communications over heterogeneous, time-varying, multiuser and interactive networks. It discusses a continuously rate-scalable foveated image and video coding system that has a number of good features in favor of network visual communications. The foveation approaches proposed in the literature may be roughly classified into three categories: geometric method, filtering-based method, and multiresolution method. The general idea of the geometric methods is to make use of the foveated retinal sampling geometry.
