The optical flow calculation is one of the fundamental problems in the field of computer vision. It serves as the basis for a wide range of applications, such as: motion detection, object and people tracking, three-dimensional reconstruction of scenes, collision detection and estimation, medical image analysis, security and surveillance, satellite image processing, robot guidance, etc. In recent years, our research group has compiled significant experience and has made important contributions in this area.

The objective of this project is to investigate on new methods for the estimation of the optical flow which are more accurate and faster. The methods we have proposed so far are part of variational techniques that produce systems of partial differential equations. These techniques have some features that make them especially useful for optical flow calculation, in fact, the most accurate methods that are known so far are related to these techniques. This project is structured as a continuation of a previous project that have been very successful and have opened new lines of research. With this project we will explore some important ideas, such as: the study of methods that exploit redundant information between frames which are not consecutive, in order to obtain more robust and accurate solutions; the inclusion of temporal information in such a way that allows us to conveniently treat the occlusions that occur in the sequence; the application of advanced strategies to accelerate the numerical schemes that are generated. Within these strategies, we consider the use of specialized hardware for the implementation of our methods, such as graphics processing units (GPU), the use of libraries for the parallelization of algorithms or more advanced numerical realizations.