Video-based Salient Object Detection: A Survey
Vivek Kumar Singh, Parma Nand, Pankaj Kumar Sharma, Preeti Kaushik · 2022 International Conference on Computing, Communication, and Intelligent Systems (ICCCIS) · 2022
Salient object detection aims to detect most important regions from the natural visual scenes. Saliency detection task has attracted great attention of computer vision researchers in static visual scenes due to its significant applicability and challenging problem. Recently, salient object detection has been explored in understanding and modelling significant regions detection over dynamic scenes. This salient object detection is referred to as video-based salient object detection. In this direction, several salient object detection models have been proposed in the literature which are focused on characteristics of real dynamic scenes. Video-based salient object detection model finds the motion-related salient regions in dynamic scenes. These approaches take the motion feature and spatiotemporal constraint jointly from video sequences for saliency detection. Toward this end, this study introduces a review on video-based salient object detection. This review mainly focuses on recent progress in salient object detection in dynamic scenes. In this study, we provide a comprehensive review that presents various most challenging scenarios of salient object detection approaches in dynamic scenes and summarizes the more challenging problems of the existing methods. Further, we discuss the issues in video-based saliency detection and provide future research directions. Moreover, the study and evaluation of various benchmark video saliency detection datasets and evaluation measures are briefly presented.