Effect-Selection Tool using Visual Saliency Maps and its Evaluations
Natsumi Suzuki, Yohei Nakada · 2018
This paper proposes an effect-selection tool using a visual saliency map to improve the visual appeal of an object in a target image. In the proposed tool, an optimization problem is considered that constructs an effect sequence for editing the target image. Here, "effect" means an editing process such as a blur effect on the background of the target image. To construct the effect sequence, this tool considers the effect set that comprises effects frequently used by image designers and/or image editors. This optimization problem is formulated with an objective function calculated via a saliency map and a constraint condition concerning the image distance between the original image and the edited image. A heuristic method based on a greedy algorithm is applied to obtain an approximate solution. Thus, the proposed tool can improve the visual appeal of the target object without changing the original image into an unacceptably edited image in reasonable computational time. For validation of the proposed tool, evaluations were conducted via a questionnaire survey and time measurement.