Based on scaling laws describing the statistical structure of turbulent motion across scales, we propose a multiscale and non-parametric regularizer for the estimation of velocity fields of  bidimensional or quasi-bidimensional flows from image sequences. The spatial regularization principle used in order to close the ill-posed nature of motion estimation is achieved by constraining motion increments  to behave through scales as the most likely self-similar process given some image data. Motion estimation is formulated as a minimization problem where the solution  second order structure function is constrained to behave as a power law. The most likely power law given the image data is jointly inferred by maximization of Bayesian evidence. The motion estimator accuracy is first evaluated on a synthetic image sequence of simulated bidimensional turbulence. Results obtained with the approach based on the physics exceeds the best state of the art results. Then, the method is used to analyze a real meteorological image sequence. Selecting from images the most evident multiscale motion model enables the recovery of physical quantities which are of major interest for turbulence characterization. In particular, from the meteorological images we are able to estimate the energy and enstrophy flux, which are in agreement with previous in situ measurements, and to reconstruct structure functions and the energy spectrum with information of the turbulent cascades.