Automated parameter tuning for tone mapping using visual saliency

While many tone mapping operators have been proposed, tuning their parameters remains a challenging task. In this work, we present a parameter tuning algorithm that can optimize the parameters of tone mapping operators automatically by minimizing the distortion in visual saliency caused by the process of tone mapping. The algorithm employs an improved saliency detection model for HDR images. Saliency distortion is quantified as the Kullback-Leibler divergence between the saliency distributions of the tone mapped images and those of the corresponding HDR images. We show that the minimization can be accomplished by employing an evolution strategy. Experiments using several tone mapping operators and a number of HDR images demonstrate the effectiveness of our algorithm. Statistical analyses are conducted to assess the improvement over default parameter settings and prior work. Graphical abstractDisplay Omitted HighlightsWe provide a universal solution for parameter tuning of tone mapping operators without the requirement for user interaction.We use the bottom-up visual saliency calculated from computational visual attention systems to measure image quality for parameter tuning.We employ an evolution strategy to solve the automatic parameter tuning as an optimization problem.Our method can improve the image quality of tone mapped images by generating results with more faithful appearance.