Research And Implementation Of Mesh-Based Motion Estimation And Compensation

Motion estimation/motion compensation (ME/MC) is a key technique in image video compressing. The wavelet is first used in still-image compression since its advantages in the image processing. Prompted by the promising performance of recent wavelet-based still-image compression algorithms, there has recently been interest in deploying ME/MC to produce wavelet-based video coders. However, the fact that the usual critically sampled DWT used in image-compression efforts is shift variant greatly hinders the ME/MC process when deployed in wavelet domain. In order to overcome the shift variance of DWT, a number of proposals have been made to use an overcomplete, or redundant, wavelet transform for ME/MC since such a redundant discrete wavelet transform (RDWT) lacks subsampling and is thus shift invariant. This dissertation will consequently explore the use of RDWT in video coding.Irregular mesh is more approximate the texture and the profile of the motion object. In this paper, a new motion estimation and motion compensation scheme based on irregular mesh is proposed. The choosing of proper control points and PMA (Potential Motion Areas) is completed using a shift invariant redundant wavelet transform, the motion estimation/compensation takes place through an affine transformation in spatial/time domain. Furthermore, based on the redundant transformation, we find a mask of the PMA.The images are selected from the international standard test warehouse, and are QCIF/CIF (YUV). The luminance component is used in the experiments. The experimental results presented in this paper show that the proposed method outperforms the other two motion estimation/compensation operations in both the spatial and redundant wave domains.