| 摘要 |
<p>The invention concerns a method for evaluating the quality of audio-visual sequences by: a training comprising the attribution of a subjective rating to each of No training sequences exhibiting degradations identified by a training vector assigned to each sequence according to a first vectoring process, to constitute a database consisting of No training vectors MOi and subjective ratings NSi; classifying the training vectors into k classes of ratings based on the subjective ratings NSi which have been attributed, to form k training sets whereto are attributed k significant ratings; establishing for said audio-visual sequence to be evaluated a vector according to said first vectoring process; attributing to the audio-visual sequence the significant training rating NSRj, corresponding to the closest training set Eaj.</p> |
