Selective attention, ideal observer theory and 'early' visual channels

Most traditional models of early vision assumed fixed channelling of information through the visual system. However, recent thinking has emphasized the flexibility of the processing and readout of information coming through the visual system, depending on such factors as attention, task demands, learning and practice (see Julesz and Kovacs, 1995; Nayar and Poggio, 1996; Karni and Bertini, 1997). Here, we propose a two-stage model based on these new principles. It was developed to explain our empirical results on binocular summation (Pardhan and Rose, 1999). Although qualitative at present, we believe it provides a framework for developing future quantitative, integrative models of channelling in binocular vision.