Abstract

This paper reports the results of a side-by-side evaluation of nine candidate metrics for quantifying edge blend quality in multi-channel display systems. The metrics tested include eight variations of contemporary luminance difference and slope-based metrics and a new approach based on a just-noticeable-difference area (JNDA) analysis. The performance of each candidate metric was evaluated by applying each metric to a set of 135 edge blended images which had been degraded using blend zone perturbations typical of those found in multi-channel flight simulators. The pool of degraded images was evaluated by 16 observers who produced ratings of edge blend quality for each image. The performance of each metric was evaluated by calculating the correlation between the metrics and the ratings of quality. The correlation between the JNDA metric and quality ratings was 0.84 and JNDA metric clearly out performed the competing candidate metrics.