Subjective quality evaluation of different intraframe adaptive coding schemes, based on orthogonal transformations

This paper describes a subjective test for comparing different video intraframe coding methods based on adaptive orthogonal transformations. A set of images were objectively measured, in terms of mean square error, by means of computer simulation using a 16×16 Discrete Cosine Transform (DCT) in the following situations: - non adaptive transform coding - adaptive coding obtained through multiplicative coefficient for the quantiser level, according to the activity classes - adaptive coding with a constant mean square error per sub-block. Subjective and objective measures show a low correlation if a non adaptive coding scheme is used, while the correlation is higher if adaptive coding schemes are used. In particular, the uniform impairment distribution (the same quantisation error for each block) supplies a very high correlation between the two measures. Subjective evaluations were carried out using a SSR (Single Stimulus Rating) method; results are presented in terms of mean and standard deviation. Analysis of variance (ANOVA) was also performed to evaluate factor and interaction contribution to variability.