Nonlinear issues complicate the modeling of the user-perceived quality of compressed images, especially when setting up feature-based objective methods. The research described in this paper indicates that the co-occurrence matrix and its features can be effective descriptors, embedding relevant information with respect to the perceptual phenomenon of visual quality. The paper first proposes a statistical approach to correlating objective quantities with subjective empirical evidence, then applies such a protocol to features associated with co-occurrence matrixes. Models from Computational Intelligence eventually support the actual quality-prediction task. Experimental results on real benchmarks confirm the general validity of the approach.
Co-occurrence matrixes for the quality assessment of coded images
GASTALDO, PAOLO;ZUNINO, RODOLFO
2007-01-01
Abstract
Nonlinear issues complicate the modeling of the user-perceived quality of compressed images, especially when setting up feature-based objective methods. The research described in this paper indicates that the co-occurrence matrix and its features can be effective descriptors, embedding relevant information with respect to the perceptual phenomenon of visual quality. The paper first proposes a statistical approach to correlating objective quantities with subjective empirical evidence, then applies such a protocol to features associated with co-occurrence matrixes. Models from Computational Intelligence eventually support the actual quality-prediction task. Experimental results on real benchmarks confirm the general validity of the approach.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.