New methods for the computation of optical flow

Various applications of partial differential equations in image processing were studied and applied to medical images. The problem of the computation of optical flow was analyzed. A novel approach defining an extended optical flow constraint was presented. The new terms included follow the idea of the conservation of weighted areas.; For determination of optical flow from a sequence of images a new method based on total variation was next presented. It was shown that the approach has inherent advantages since it does not smooth the flow-velocity across the edges and hence preserves edge information. A numerical approach based on computation of evolving curves was proposed for computing the optical flow field. The above methods were combined to create a robust approach to computing optical flow.; Scale-space multi-resolution methods were implemented to create a rapid mechanism for numerical implementation of the computations. The multigrid methods used the residual to smooth and decimate repeatedly while moving to coarser grids upon which to solve the equations. Finally a correction interpolation scheme was used to repeatedly relax back to the fine grid and the solution.; Computations were carried out on a number of real and synthetic image sequences in order to illustrate the theories as well as the numerical approaches. Comparisons were made with existing techniques as to the accuracy of the computed flow fields using synthetic test image sequences. The new methods showed improved calculations. The vector fields were also computed for a variety of medical imaging applications, particularly cardiac images in healthy and pathological states. It was found that the modified methods better captured edge information and global changes in optical flow. The methods were shown to be useful as an aid to describe cardiac motion and edge information in images.