Digital video communication has evolved tremendously§in the past few years, experiencing significant§advances in compression and transmission techniques.§To quantify the performance of a video system, it is§important to measure the quality of the video. Since§humans are the ultimate receivers of a video signal,§quality metrics must take into account the properties§of the human visual system. So far, most of the§metrics that have been proposed require access to the§original video, what makes them unsuitable for§real-time applications. We investigate how to§estimate video quality in real-time applications§using no-reference and reduced reference metrics. For§this, we study the visibility, annoyance, and§relative importance of different types of artifacts§and how they combine to produce annoyance. The work§uses synthetic artifacts that are simpler, purer, and§easier to describe, allowing a high degree of control§with respect to the amplitude, distribution, and§mixture of different types of artifacts. We present§metrics for estimating the strength of four types of§artifacts. The outputs of the best artifact metrics§are used to build a combination model for overall§annoyance.