Low Bit-Rate Video Coding Using Iterative Affine Motion Estimation and Quadtree Segmentation 1
|
Abstract:
In this work we present a low bit-rate hybrid coder which achieves a significant rate reduction, as compared to the standard H.261 coder, through adaptive motion segmentation, improved motion compensation, and efficient representation of motion-model parameters. In the proposed coder, each frame is adaptively segmented into non-fixed-size blocks using a quadtree approach by which a block is split if the estimated motion parameters for that block do not provide adequate motion compensation. Motion estimation is performed by an Iterative Least Squares algorithm using an affine motion model for large blocks and the common translation motion model for the minimum size blocks. An efficient representation of the motion-model parameters is obtained via 'corner vectors' which are shown to provide reduced sensitivity to quantization errors. Simulation results show a particular advantage of the proposed coder at bit rates of 64Kbps and below. 1
