This is chapter 4 of the textbook Understanding Vision: theory, models, and data Oxford University Press, 2014.
Abstract: This chapter discusses the difficulties in understanding the primary visual cortex (V1) by the role of efficient encoding of visual information. It shows that the redundancy in visual input is mainly in the second order, rather than the higher order, statistics of visual inputs, and that the higher order, rather than the second order, redundancy contains much of the useful information about visual objects in the scene. Since much of the second order redundancy is removed by the efficient visual encoding in the retina, this chapter argues that that V1, the largest cortical area in the brain, with an overcomplete representation of visual inputs in its neural responses, is unlikely to be devoted to further increasing the encoding efficiency by removing the higher order redundancy. It thus argues that V1 must be serving some other important function in vision.
Keywords: overcomplete representation, efficient coding, primary visual cortex
Its figures in a pptx file