Bandlets for HDTV Video Processing
Stéphane Mallat, Joan Bruna · 2008
Summary form only given. De-interlacing, scaling, frame rate conversion, compression artifact removals are currently major video processing challenges to produce high quality images for HD TV displays. Recovering high resolution images from degraded or lower resolution videos requires adaptive geometric filtering and interpolation in space and time, at very high data rate. Finding appropriate image representations for geometric image processing is currently a fundamental research problem. We first review some psychophysic and physiologic models of human geometric visual processing. We then introduce bandlet bases, which are multi-scale geometric wavelets, constructed along space-time trajectories adapted to the video content. Applications to HDTV video processing will be shown.