Towards the design of a consistent image contrast enhancement evaluation measure

Contrast Enhancement Evaluation (CEE) is a very challenging problem. In this work, we provide a detailed performance analysis of CEE measures. The study was conducted on our newly developed database dedicated to psychophysical Contrast Enhancement (CE) quality evaluation. The database contains 30 original images and 180 enhanced images obtained using six different CE methods as a representative set of the most common approaches used in the literature. The CE methods were subjectively evaluated and ranked by 23 observers using a Pairwise Comparison (PC) protocol. The correlation analysis between the subjective preferences and objective evaluations of the enhanced images show that most of the existing CEE metrics are not well consistent with the human judgment of quality. We also present in this paper a thorough discussion on the available CEE metrics, their strengths, their weaknesses, and their inter-correlation. In addition to the individual metrics, we show that by fusing different metrics together, a significant increase in correlation performance can be achieved. This study reveals that there is a clear need to develop more robust CEE measures which are perceptually motivated and correlated well with the quality of enhancement a given image is subjected to. The new database introduced in this paper is expected to contribute substantially promoting such research area, which is of primary importance to diverse multimedia applications.