Low-power uniform video acquisition with super-resolution reconstruction

A content-based video acquisition system was proposed recently for the power optimization of portable digital devices in. In this scheme, extra computation is required in the pre-analyzing for detecting the spatially and temporally active regions. Moreover, given full-resolution frames (FR) at certain frequency for reference, the adaptive sampling at the high active regions is redundant once proper super-resolution is applied. We propose in this paper a low-power video acquisition system by capturing frames in groups, with one FR frame and one low-resolution (LR) frame in each group. Consequently, the prerequisite of content-based video compression in is waived and the acquired video can be processed by any video compression techniques. The proposed scheme can reduce the power consumption by 37.5%, higher than in most situations. The acquired video is post-processed by an example-based superresolution (SR) technique to recover the FR. Experiments show that the reconstructed FR video can achieve better objective and subjective quality than the smart sampling in.