Search
Login »
 
Publication details
Authors F.BartoliniA.Piva
Title Median based relaxation of smoothness constraints in optic flow computation
Abstract In this work, a nonlinear filtering technique is used to enhance the performance of the classical Horn and Schunck global optimization algorithm for optic flow computation. The Horn and Schunck algorithm is an iterative procedure which refines, step by step, the optic flow estimate. At each step the optic flow field is low pass filtered by a linear operator. This filtering operation produces, as a side effect, an oversmoothing of the resulting optic flow field. This effect is particularly annoying across motion boundaries. In this letter it is proposed to substitute a median (non-linear) filter for the linear operator, in order to reduce the strength of the smoothing step, especially across motion boundaries.
Publication Pattern Recognition Letters
Typology International Journal Papers
Date July 1997
Other info vol. 18, no. 7, pp. 649-655, 
 
Article Download »
  
Powered by Trimatica s.r.l.