Method of evaluating the quality of audio-visual sequences

摘要

The invention provides a method of evaluating the quality of an audiovisual sequence by: a) training, comprising allocating a subjective score NS_i to each of N₀ training sequences S_i (where i=1, 2, . . . , N₀) presenting degradations identified by a training vector MO₁ which is given to each sequence S_i in application of a first vectorizing method, in order to build up a database of N₀ training vectors MO_i with subjective scores NS₁; b) classifying the N₀ training vectors MO_i into k classes of scores as a function of the subjective scores NS₁ that have been allocated to them, so as to form k training sets EA (where j=1, 2, . . . , k) which have k significant training scores NSR_j allocated thereto; c) for each audiovisual sequence to be evaluated, generating a vector MO using said first vectorization method; and d) allocating to the audiovisual sequence for evaluation the significant training score NSR_j that corresponds to the closest training set EA_j.