Abstract:

    In traditional methods, when the relative motion between video objects in consecutive instants was complex, only the spatial shape error concealment was used with not very desirable effect. In order to obtain better concealment results, a Temporal shape error concealment scheme Robust to Rotation and Translation (TRRT) is proposed. This scheme is based on the invariant properties of rotation and translation of both Harris interest point detector and local Zernike moments. First, the best matching pairs of interest points are obtained by comparing the Euclidean distance of local Zernike moments defined on the neighborhood of the interest point. Then, the global motion parameters between video objects are determined and motion compensation is given to the reference boundary. Finally, the missing boundary pieces are reconstructed by boundary matching and compensation. This new scheme is compared with traditional temporal and spatial methods. Results show that better concealment can be obtained by this new scheme.

Key words: information processing, shape error concealment, motion compensation, interest point, local Zernike moments