Subjective testing has long been used to quantify user preference in the field of imaging. The majority of subjective testing is done to analyze still images, leaving the ever-growing field of video overlooked. With little work put into this area of study, not much is known about the preferential behavior of dynamic auto control functions such as automatic exposure (AE). In this study, we focus on subjective preferences for two aspects of video auto exposure convergence: convergence time and convergence curve type, with each tested individually. This experiment utilizes a novel framework for subjective testing, where a collection of videos are captured with simulated changes in light. This method allows for much more precise control of the capture device and constitutes better repeatability of experiments, as opposed to recording real changes. A paired comparison model is employed to conduct the subjective analysis of the videos. In a web application, two videos are played side by side with a slight delay and the user is asked to pick which video they prefer. Results from the experiments show that users prefer monotonic, gradual transition in AE, with no sharp or abrupt changes. Users also preferred transition times of 266-500 milliseconds.
Seungseok Oh, Clayton Passmore, Bobby Gold, Taylor Skilling, Sean Pieper, Taek Kim, Margaret Belska, "A Framework for Auto-exposure Subjective Comparison" in Proc. IS&T Int’l. Symp. on Electronic Imaging: Image Quality and System Performance XIV, 2017, pp 202 - 208, https://doi.org/10.2352/ISSN.2470-1173.2017.12.IQSP-244