基于区域对比的图像显著目标检测

Image Salient Objects Detection Based on Region Contrast

针对具有杂乱背景图像的显著目标检测问题,提出了一种无需任何先验知识,通过分析计算区域平均显著值的对比度来提取显著目标的方法.根据显著图,计算出显著目标的最小边界框与其周围区域的显著性差异,且通过折半查找的方法加快搜索的收敛速度.结果表明,该算法能迅速检测出与人类视觉感知结果一致性较高的显著目标区域.

To handle the salient object detection in images with cluttered backgrounds,a new method to identify salient object without any prior knowledge was proposed.It calculates and maximizes the region contrast between the target regions and their surroundings by a binary search to accelerate the convergence.Results showed that the proposed algorithm could detect the salient object region consistent with human visual perception system in a faster way.