Decoding MT Motion Response for Optical Flow Estimation: An Experimental Evaluation

N V Kartheek Medathati (1) , Manuela Chessa (2) , Guillaume S. Masson (3) , Pierre Kornprobst (1) , Fabio Solari (2)

1 NEUROMATHCOMP - Mathematical and Computational Neuroscience
2 DIBRIS - Dipartimento di Informatica, Bioingegneria, Robotica e Ingegneria dei Sistemi [Genova]
3 INT - Institut de Neurosciences de la Timone

Motion processing in primates is an intensely studied problem in visual neurosciences and after more than two decades of research, representation of motion in terms of motion energies computed by V1-MT feedforward interactions remains a strong hypothesis. Thus, decoding the motion energies is of natural interest for developing biologically inspired computer vision algorithms for dense optical flow estimation. Here, we address this problem by evaluating four strategies for motion decoding: intersection of constraints, maximum likelihood, linear regression on MT responses and neural network based regression using multi scale-features. We characterize the performances and the current limitations of the different strategies, in terms of recovering dense flow estimation using Middlebury benchmark dataset widely used in computer vision, and we highlight key aspects for future developments.

Mots clés: spatio-temporal filters; motion energy; Bio-inspired approach; optical flow; MT; Middlebury dataset; V1; population code

Domaines: Intelligence artificielle [cs.AI]; Neurosciences; Informatique; Bibliothèque électronique [cs.DL]; Recherche d'information [cs.IR]

N V Kartheek Medathati, Manuela Chessa, Guillaume S. Masson, Pierre Kornprobst, Fabio Solari. Decoding MT Motion Response for Optical Flow Estimation: An Experimental Evaluation. [Research Report] RR-8696, INRIA Sophia-Antipolis, France; University of Genoa, Genoa, Italy; INT la Timone, Marseille, France; INRIA. 2015. ⟨hal-01131100⟩