Low-Power Content-Based Video Acquisition for Super-Resolution Enhancement

Motivated by the optimization of power and improvement of video resolution, this paper proposes a content-based adaptive sampling system for video acquisition. Blind sampling suffers from the lack of resolution and blurring. However, use of a priori knowledge can provide intelligent sampling function that will reduce the blur artifacts. This paper proposes an information theoretic criteria-based sampling function. Higher sampling is proposed at high motion and edge regions while lower sampling at the low-frequency regions. This helps in providing better resolution with lower power consumption. Previous researches focus on enhancing the coding performance after the video acquisition stage. The proposed adaptive sampling scheme naturally performs super resolution without requiring extensive postprocessing. The proposed scheme has been tested on ten exemplary video sequences. Quality of the proposed adaptive sampling method is over 10-16 dB better than the coarsely sampled video. The power savings is ap 30-40% compared to acquiring the full resolution video.