Weighting coefficient determination in a displacement estimation functional using the dynamical similarity principle

In the unconstrained minimization techniques, the determination of weighting coefficients values is always a difficulty. This is the reason why the majority of previous works about displacement estimations with a functional, assume an experimental determination of the coefficients values with few transfer possibilities to other experiments. A method of reducing the difficulty is to define the functional from physical properties rather than geometrical considerations as "smoothness constraint". The introduction of a physical model reduces the admissible solutions set and give a physical signification to the coefficients and then a dimension. Starting from this consideration, we present a method based on the dynamical similarity to transfer the weighting coefficients values from a sequence of images to an other one.