Object-Based Very Low Bit-Rate Coding Using Motion Parameter Estimation Based on Multiple Frame Prediction

Abstract

This paper proposes an object-oriented coding method in very low bit-rate channels, using the motion estimation technique based on multiple frame prediction. In most existing methods, only the global motion components are transmitted. So the images containing abrupt motions and noise yield large error, even after texture compensation using the discrete cosine transform (DCT). To reduce the global motion prediction error due to usage of rigid motion parameters, residual image mapping is employed. Since high computational complexity is required for local motion correction using residual image mapping, simplified global motion prediction prior to local motion refinement parameter extraction is proposed to reduce the burden of the overall coder complexity. Computer simulation results show that the proposed method gives good performances in terms of the subjective quality, the peak signal to noise ratio (PSNR), and the compression ratio (CR).