Video quality is the degree at which a given video suits the end-user's expectations. It is therefore a subjective notion. The proof is that a given video, viewed in the same conditions by different observers, can be judged very differently by these observers. One observer could say that the video quality is good while another one could say that the video quality is bad, and both could be right: because it depends on the video quality they expect and on the video quality they're used to have.
Video quality can be expressed using another way: indeed, video quality can be measured by measuring visual annoyance because visual annoyance is often a bit easier to evaluate than video quality. Visual annoyance represents the level of difficulty you meet when trying to perform a given task (from "simple comfort" to face expressions detection or object recognition) in presence of video distortions (like encoding artifacts or decoding errors due to packets loss during transmission).
Video quality judgment and visual annoyance both come from visual perception and can be represented by the top-down approach below:
Video quality is a subjective notion and therefore the best way to measure video quality is based on human observers. This way is called "subjective video quality assessment". It requires to use human observers who must score the video quality of videos they're watching during experiments called "video quality assessment tests".
However, in order to collect precise video quality measures, many recommendations have to be followed when realizing such subjective video quality assessment tests. Otherwise, the collected video quality scores often become useless (due to the lack of precision on the video quality scores).
These recommendations lead the subjective video quality assessment method to be very complicated. More, performing subjective video quality assessments tests is a long and expensive process. To get more information on this topic, please read AccepTV's white paper on subjective video quality assessment methods.
Video quality is difficult to measure with the "subjective video quality assessment" method descrived above. Luckily, there is another way to measure video quality. This other way is called "objective video quality measurement".
It is based on the use of a computational method, an algorithm, called "metric" (or "video quality metric" or "video quality criterion") which produces values expressing video quality.
One of the fundamental properties required for a video quality metric is that it should produce objective video quality scores which are well correlated with subjective video quality scores collected from human observers during video quality assessment tests. From a practical point of view, a video quality metric is an algorithm able to score (on a scale) the video quality of a tested video which may have been distorted (usually due to encoding and/or transmission).
While computing video quality scores can be very easy, producing precise, meaningful and coherent video quality scores (which means video quality scores well correlated with subjective video quality scores given by human observers) is much more complicated and requires important research efforts on human visual perception, features extraction, distortions visibility and pooling of disortions into video quality scores.
The interest of these video quality metrics is that they enable to measure or monitor in real time the video quality of audio video services, in a fully automatic and repeatable manner. While video quality measurement is often used for equipment benchmarking (to buy the best equipment, the one which provides the best video quality), video quality monitoring solutions can trigger alerts when equipment fails to reach a given video quality level.
To measure and monitor perceived video quality, AccepTV has developped several products: