Early investigations in optical flow from colour images

This thesis presents five new methods for optical flow recovery, all utilising colour as well as intensity information from image sequences. Most work in the field of motion estimation in the past has used intensity information only. The new methods are shown to perform well, some of them outperforming common intensity-only methods. During this research an application was created to enable the simple comparison of various optical flow techniques. It features custom image sequences, synthetic ground-truth creation, custom pre-filtering, differentiation and optical flow methods (both grayscale and colour) and flow analysis through image reconstruction, density/sparsity analysis, ground-truth analysis and flow visualisation.