MULTISCALE MODELING AND ESTIMATION OF MOTION FIELDS FOR VIDEO CODING

We present a systematic approach to forward-motion-compensated predict ive video coding. The first step is the definition of a flexible model that compactly represents motion fields, The inhomogeneity and spatia l coherence properties of motion fields are captured using linear mult iscale models. One possible design is based on linear finite elements and yields a multiscale extension of the triangle motion compensation (TMC) method, The second step is the choice of a computational techniq ue that identifies the coefficients of the linear model, We study a mo dified optical how technique and minimize a cost function closely rela ted to Horn and Schunck's criterion, The cost function balances accura cy and complexity of the motion-compensated predictor and is viewed as a measure of goodness of the motion field, It determines not only the coefficients of the model, but also the quantization method. We formu late the estimation and quantization problems jointly as a discrete op timization problem and solve it using a fast multiscale relaxation alg orithm, A hierarchical extension of the algorithm allows proper handli ng of large displacements, Simulations on a variety of video sequences have produced improvements over TMC and over the half-pel-accuracy, f ull-search block matching algorithm, in excess of 0.5 dB in average, T he results are visually superior as well, In particular, the reconstru cted video is entirely free of blocking artifacts.