Research On The Cube-based Compressed Video Sensing

Compressed sensing(CS) breaks the limits of Nyquist sampling rate and achieves a direct sampling for information. Compressed video sensing(CVS) introduces CS into video encoding and decoding, and thus realizes a new method of encoding.The existed compressed video sensing schemes are mostly block-based or frame-based encoding and decoding the video signal. However, for three-dimensional video data, the cube-based compressed video sensing scheme would ideally be a more intuitive method in the sense that CS measurements would span the entire spatial and temporal extent of a video sequence. Unfortunately, such global CS acquisition of video requires a complex and expensive spatial-temporal-light modulation, and thus is largely considered impractical to implement in a real device. In this paper, a novel compressed video sensing scheme which is exactly suitable for wireless multimedia sensor networks is proposed by changing the method of processing multiple frames in video cubes. At the encoder, the scheme adopts frame-based measurement with the low complexity to well adapt to the wireless network with the energy-limited underlying nodes. And it enhances the decoding quality of the first and last frame by reallocation the measurements to increase the measurements contained in them. What’s more, by assisting the reconstruction of intermediate frames with the first and last frames, their decoding quality is also enhanced and the performance of the whole system is improved. At the decoder, the measurements are similar to measurements of the global CS acquisition by global scrambling and thus improving the performance at low sampling rate. Simulation results show that the proposed scheme has a high quality decoding performance, especially in the case of low sampling rate. The proposed scheme can yield about 4~5d B improvement in decoding quality compared with the frame-wise measurement stacking reconstruction(FS). There is 0.4~0.9d B increase compared with the multi-frame measurement multi-frame reconstruction(M3R). The result show the proposed scheme effectively improves the decoding performance without impacting realizable hardware devices.