Computational analysis of disparity modulation sensitivity: an explanation in terms of a Bayesian surface reconstruction

Sensitivity to binocular disparity modulation has been shown to have a bandpass nature. This paper presents a computational account for the disparity modulation function (DMF) in terms of a Bayesian surface reconstruction. The Bayesian approach suggests that prior assumptions about surface structure will affect the perception of disparity modulation. Taking into account a prior constraint of surface smoothness being imposed on the perceived surfaces, we propose that computing the first derivatives of the surfaces determines the bandpass shape of the DMF. Based on this idea, we derive an analytical prediction of the DMF. It is then shown that the prediction gives a good fit to the empirical data. Implications for possible mechanisms underlying the DMF are discussed.