Our parametric metrics are precise and fast. In fact they don't really take more time than video decoding.
Why? Because actually they catch up information from the video decoder and use it to compute the video quality score.
For example, these metrics get the type of frame (I, B, P), the QP (quantization parameter) value of each macro-block, the motion vector of each macro-block, the decoding status of each macroblock (correctly decoded or badly decoded due to packet loss).
The main interest is that the decoder can provide a very precise information about the decoded image. For example, the decoder knows exactly which blocks were correctly decoded and which blocks were not (and why). All the collected information is then injected in a Human Vision model which computes a video quality score.
These parametric are specially interesting when measuring the impact of transmission errors on perceived video quality. Opposite to QoS (Quality of Service) metrics that measure a packet loss rate but don't know the impact of each lost packet on video quality our parametric solutions DO decode the video and KNOW the impact of a lost packet on QoE.
So our parametric metrics know if some macroblocks couldn't be correctly decoded. For each of these macroblocks, they know which type of information is missing: motion vector, residue, etc. And they know where these macroblocks are in the image. Therefore, transmission errors are detected and their impact on perceived quality is measured.
We have developed parametric metrics for the two major codecs: