A unified random field model based neural approach to pixel labeling problems in computer vision
G. Parthasarathy, P. Ananth Raj · 1994
In this paper, we establish a nexus between the Gibbs random field models and recurrent neural networks and present a unified neural network approach based on this to pixel labeling problems in computer vision. We also present a case study with photometric stereo problem for demonstrating the viability of the present approach.>