Motion Estimation in Measurement Domain for Compressed Video Sensing

Compressed sensing (CS) is a new approach to signal acquisition that can potentially allow us to design very simple video encoders, which then develops as the compressed video sensing (CVS). However, a unique characteristic of CS is that it directly captures the signal in the measurement domain, which then makes these traditional techniques which are used to remove the time redundancy of video signals, as motion estimation, not suitable for compressively sampled videos. In this paper, by analyzing the inner spatial relation in original signals and relationship between original signals and measurements, a new motion estimation method in measurement domain for CVS is proposed. By using this method, we can perform motion estimation directly with the sampled measurements, thus offering convenience to remove the time redundancy of videos directly in measurement domain.