A human visual system model for no-reference digital video quality estimation

Francesco Massidda; Cristian Perra; Daniele D. Giusto

doi:10.1117/12.594039

Translator Disclaimer

17 January 2005 A human visual system model for no-reference digital video quality estimation

Francesco Massidda, Cristian Perra, Daniele D. Giusto

Author Affiliations +

Proceedings Volume 5668, Image Quality and System Performance II; (2005) https://doi.org/10.1117/12.594039
Event: Electronic Imaging 2005, 2005, San Jose, California, United States

Abstract

No-reference metrics are very useful for In-Service streaming applications. In this paper a blind measure for video quality assessment is presented. The proposed approach takes into account HVS Luminance Masking, Contrast Sensitivity and Temporal Masking. Video distortion level is then computed evaluating blockiness, blurring and moving artifacts. A global quality index is obtained using a multi-dimensional pooling algorithm (block, temporal window, frame, and sequence levels). Different video standard and several compression ratios have been used. A non-linear regression method has been derived, in order to obtain high linear and rank order correlation factors between human observer ratings and the proposed HVS-based index. Validation tests have been developed to assess index performance and computational complexity. Experimental results show that high correlation factors are obtained using the HVS models.

Citation Download Citation

Francesco Massidda, Cristian Perra, and Daniele D. Giusto "A human visual system model for no-reference digital video quality estimation", Proc. SPIE 5668, Image Quality and System Performance II, (17 January 2005); https://doi.org/10.1117/12.594039

ACCESS THE FULL ARTICLE