Discrete Orthogonal Decomposition and Variational Fluid Flow Estimation
The decomposition of motion vector fields into components of orthogonal subspaces is an important representation for both the analysis and the variational estimation of complex motions. Common finite differencing or finite element methods, however, do not preserve the basic identities of vector analysis. Therefore, we introduce in this paper the mimetic finite difference method for the estimation of fluid flows from image sequences. Using this discrete setting, we represent the motion components directly in terms of potential functions which are useful for motion pattern analysis. Additionally, we analyze well-posedness which has been lacking in previous work. Experimental results, including hard physical constraints like vanishing divergence of the flow, validate the theory.
KeywordsOrthogonal Subspace Motion Vector Field Subspace Correction Real Image Sequence Scale Space Method
Unable to display preview. Download preview PDF.
- 13.Evans, L.C.: Partial differential equations. In: Graduate Studies in Mathematics, vol. 19. American Mathematical Society, Providence (1998)Google Scholar