Quantitative Models of Image Quality

Abstract

Fourteen image quality metrics were evaluated for hard-copy and soft-copy displays of digital images degraded by various levels of noise and blur. All quality metrics were formulated to include the displayed modulation spectrum of the image. The statistical analyses suggested that several of the metrics correlated strongly with performance, and, thus, support the proposed utility of image-dependent quality measures. An MTFA-type metric was shown to correlate highest with the average performance scores across noise and blur conditions.