In this thesis, two topics in video compression are investigated: Mesh based motion estimation and compensation, and Fast mesh based motion estimation....[ Read more ]

In this thesis, two topics in video compression are investigated: Mesh based motion estimation and compensation, and Fast mesh based motion estimation.

Mesh based motion estimation generates smooth motion field and good visual quality without blocking artifacts. However, the search for grid point locations is computation intensive due to the interpolation and warping. In this thesis, a fast mesh based motion estimation algorithm called FMMA is proposed. The proposed FMMA is an iterative algorithm using block-based block estimation algorithm with progressively smaller block size to estimate the motion of the grid points. Optimizations of FMMA are also discussed.

Experimental results show that FMMA can achieve significant computation reduction at the expense of minor visual degradation comparing to traditional MMA. Finally optimizations of FMMA are discussed and they outperform traditional MMA by a factor of about 340.