融合图的流形排序与引导学习的显著性目标检测

摘要

为了在显著性目标检测中保持高的召回率的同时提高准确率,本文提出了3点改进思路.第一,从超像素中提取简单的视觉特征,如颜色、方向和空间信息;第二,为了克服经典的基于图的流形排序(MR)的显著性目标检测算法中背景先验假设的缺点,使用仿射传导聚类算法(APC)自动聚合超像素为不同的特征类别.根据目标与背景(改进的)边缘连通度的不同,图像边缘的超像素会得到较大的权重即较大的背景概率值,这样边缘上真正的背景超像素就会筛选出来.同时,使用改进的MR算法计算图像的显著性值.第三,为了进一步增强算法的性能,前面第二步的结果可以作为“弱”显著图,利用引导学习算法从中产生“强”显著图得出最终结果.基于3个标准图像库的实验结果证明,本文提出的算法在性能上超过了其它3种优秀算法.%To increase precision while preserving high recall in salient object detection,three schemes have been proposed.First,simple visual features,namely color,orientation,and spatial information,are used to represent image superpixels.Second,to overcome the shortage of the boundary prior assumption based on graph-based manifold ranking (MR),the affinity propagation clustering (APC) is utilized to aggregate the superpixels (nodes) to different feature clusters adaptively.According to the modified boundary connectivity,the superpixels along the image boundaries are assigned with different background weights (the values of background probability).The real background seeds are selected and an improved MR method is employed to compute saliency.Third,to further improve the performance,the result of the second step acts as the weak saliency map.The bootstrap learning algorithm is used to generate the strong saliency map and the final result.Comparing with other three state-of-the-art algorithms on three public benchmark datasets,our experimental results demonstrate that our approach outperforms other algorithms.