This is chapter 6 of the textbook Understanding Vision: theory, models, and data Oxford University Press, 2014, it gives a pedagogic explanation of visual inference in the brain.
Abstract: This chapter gives an account of the experimental and computational investigations in visual perception or recognition. The perceptions, including illusions, are viewed as the outcomes of inferring or decoding properties of visual scenes from the neural responses to the visual inputs. Emphasis is on understanding perception at both physiological and behavioral levels through the use of computational principles. Maximum-likelihood decoding and Bayesian decoding approaches are introduced. Examples are provided to use these approaches to understand, e.g., contrast detection, color discrimination, motion direction perception, depth illusion, and influences of context and prior experience in visual perception. Limits in the visual decoding performance, due to inefficiency in utilizing the visual input information, likely caused by the attentional bottleneck, are highlighted. Likely neural architectures to implement decoding are discussed.
Keywords: visual decoding, perception, illusion, maximum-likelihood decoding, Bayesian decoding
Its figures in a pptx file