Objective video quality assessment considering frame and display time variation

Abstract

Video quality assessment is becoming more important in current digital video application. Most of the video encoders and decoders focus in providing more visual quality in less amount of digital information, being the compression one of the most important solutions for making possible the transmission of high quality videos in the current digital transmission media. Sometimes, however, not only video compression is good enough for video quality transmission, maybe the available network bandwidth or the ability to process the video by a low-end processor is not sufficient for continuous playback. Many techniques that consider time variabilities of different sorts need to be developed to keep a digital video watchable. This work focus on a new timeline approach to compare the quality of videos for which a presentation time stamp variability is allowed. Based on video quality metrics already established the new approach can be used to help the decision about the most suitable method for real-time video transmission. Among the many possible techniques able to cope with time stamp variability, one can mention the variable delay time in video presentation, or the use of frame skipping technique, or even more compression applied to reduce the bandwidth and/or processing - improving the perceived video quality under scarce resource availability. A new timeline approach to calculate the current video quality metrics used in most of the encoders and decoders available was developed. This timeline approach takes into account not only the frames themselves, as a frame-by-frame evaluation, but both the frame and presentation time stamp are considered in the proposed approach to the problem. The quality comparison is based on a new discretization, which hopefully serves better the quality of digital video without losing perceptual quality in the process. All the work is tested in an offline solution program that is able to give numeric quality index to the tested sample videos. Those samples are selected from the current distribution of the German digital video television system, encoded in MPEG-2.