Subjective video quality metrics based on failure statistics

Subjective and Objective video quality evaluation are important problems in video processing. This work describes some new guidelines for easy and reliable determination of quality metrics that are subjectively meaningful. ‘Mean Time Between Failures’ (MTBF) and ‘Probability of Failure’ (PFAIL) are introduced as new subjective quality metrics based on failure statistics, where failure refers to errors deemed to be perceptually noticeable. These two are shown to be intuitive and scalable in the context of MPEG-2 video. MTBF is a global metric, representing how often a typical viewer observes a noticeable visual error; and PFAIL is an instantaneous metric relating to the fraction of viewers that find a given video portion to be below acceptable quality levels. These subjective quality metrics are evaluated for different video clips at bit rates in the range of 1.5 5 Mbps, and are compared with existing objective quality metrics. A desired end result of this work is the estimation of the subjective quality metrics based on objective measurements. We illustrate this using the example of an objective metric, STJM (Spatial Temporal join metric).