作者: D.A. Florencio
关键词: Video denoising 、 Rate–distortion optimization 、 Block-matching algorithm 、 Motion interpolation 、 Motion estimation 、 Video tracking 、 Artificial intelligence 、 Computer vision 、 Computer science 、 Quarter-pixel motion 、 Video quality 、 Motion compensation 、 Structure from motion
摘要: Due to the capture process, video signals are generally contaminated by noise. Furthermore, overall quality can be improved reducing sharpness in fast moving areas, except cases where eye is able track motion. We propose a relatively simple pre-processing method that address both these situations. The proposed. algorithm based on selectively removing high frequencies not well predicted motion compensation. While decision made temporal tracking, filtering done exclusively spatial domain, thus avoiding artifacts produced other methods.