基于贝叶斯框架的显著目标检测

图像和视频显著性检测在计算机视觉领域引起了极大的关注,广泛应用于物体分割、可疑物检测、图像检索等领域,在贝叶斯框架下,将人类视觉注意力理论中Bottom-Up模式被动感知与Top-Down模式主动检测按照时间先后顺序结合在一起,并结合像素的"Center-Surround"模型和核密度估计,提出一种能由粗到精逐步感知和获取视场中视觉显著性目标位置及尺度的实时显著目标检测算法,称其为基于贝叶斯框架的显著目标检测.通过在微软MSRA数据集上进行ROC和Precision-Recall测试,证明该算法取得比目前经典算法更好的效果.

Saliency Object Detection Based on Bayesian Framework

Saliency detection has gained a great deal of attention in computer vision in images and videos. It is a valuable tool in image processing,such as object segmentation,suspicious detection,image retrieval,etc. This paper proposes a saliency object detection algorithm that combines the BottomUp passive perception with the Top-Down active perception,together into Bayesian framework,detecting the salient object coarse-to-fine. The method is efficiently implemented by using the kernel density estimation and the"Center-Surround"pixel model. The ROC and Precision-Rell test result on MSRA dataset show that this method outperforms all state-of-the-art approaches.