Optical Flow and Trajectory­ Estimation Methods ­(SpringerBriefs in Computer ­Science)

Optical Flow and Trajectory Estimation Methods by Joel Gibson and Oge Marques provides an in-depth exploration of two key challenges in optical flow and trajectory estimation: finding efficient convex optimization methods for sparsity and extending sparsity to improve trajectories. The authors begin by reviewing the fundamentals of optical flow, discussing commonly used strategies and their advantages or limitations. The book then delves into the concepts associated with sparsity, including dictionaries and low-rank matrices, providing context for optical flow and trajectory methods. It covers algorithms, data sets, and performance measurement, as well as sparse regularization techniques such as total variation and robust low rank trajectories. The authors also present a novel approach using partially-overlapping patches to accelerate calculation, implemented in a coarse-to-fine strategy. Experimental results demonstrate the effectiveness of combining total variation with a sparse constraint from a learned dictionary, outperforming traditional methods. This technical brief is targeted at researchers and practitioners in engineering and computer science, offering cutting-edge insights into optical flow and trajectory estimation for both new and experienced professionals.