A perceptually relevant no-reference blockiness metric based on local image characteristics

摘要: A novel no-reference blockiness metric that provides a quantitative measure of blocking annoyance in block-based DCT coding is presented. The incorporates properties the human visual system (HVS) to improve its reliability, while additional cost introduced by HVS minimized ensure use for real-time processing. This mainly achieved calculating local pixel-based distortion artifact itself, combined with visibility means simplified model masking. overall computation efficiency and accuracy further improved including grid detector identify exact location artifacts given image. calculated only at detected averaged over all image yield an score. performance this compared existing alternatives literature shows be highly consistent subjective data reduced computational load. As such, proposed promising terms both practical reliability real-life applications.