视觉显著性检测综述

Review of Visual Saliency Detection

视觉显著性(Visual Saliency)是指人类在观察某一区域时视野中存在能够引起人类视觉关注的局部区域,该局部区域被称为显著性区域。与此对应,视觉显著性检测(Visual Saliency Detection)则主要用于凸显图像或者视频中的显著性区域。近年来,基于计算机技术的视觉显著性检测模型被广泛应用于图像分割、目标检测和视频编码等领域。因此,进行视觉显著性检测研究具有相当的实际应用价值。该文针对视觉显著性检测研究进行了系统地介绍,主要包括RGB图像显著性检测、RGBD图像显著性检测、视频显著性检测和协同显著性检测等,同时总结分析了各种视觉显著性检测模型的特点,并进一步探讨了视觉显著性检测研究未来的发展方向。

Visual saliency refers to that there is always a local area in the visual field that can arouse the most attention of the visual system when people observe a certain area, which is called the saliency area. Accordingly, visual saliency detection is mainly used to highlight saliency areas in images or videos. In recent years, visual saliency detection model based on computer technology is widely used in image segmentation, object detection and video coding and so on. Therefore, the research on visual saliency detection has considerable practical application value. This paper systematically introduces the research of visual significance detection, mainly including RGB saliency detection, RGBD saliency detection, video saliency detection and Co-saliency detection. At the same time, it summarizes and analyzes the characteristics of various visual saliency detection models, and further discusses the future development direction of visual saliency detection research.