自然场景下基于边界先验的图像显著性检测

Image Saliency Detection Based on Boundary Prior in Natural Scenes

为了对自然场景中的显著目标进行准确检测,提出一种基于边界先验的图像显著性检测方法。采用简单线性迭代聚类的超像素分割算法将图像分割为颜色和纹理具有一致性的超像素,根据边界先验理论,分别计算4个边界的边界先验显著图,并且融合成为粗略的显著图,大致区分图像的背景和显著目标,将边界先验显著图的质心作为显著目标的中心位置进行空间显著性分析,从而突出显著目标,得到最终的显著图。仿真结果表明,与Itti算法、基于对比的方法、基于图论的方法等相比,该方法能够均匀地突出显著对象,有效地抑制背景。

In order to detect saliency object accurately in natural scenes,an image saliency detection based on boundary prior in natural scenes is proposed in this paper.The original image is first segmented into a set of superpixels with similar color and texture using simple linear iterative clustering superpixel segmentation algorithm.According to the theory of boundary prior,4 boundary prior saliency maps are calculated separately.It combines them to get a coarse saliency map which separates the background and salient object roughly.The final saliency map which further highlight salient object is generated by regarding the centroid of the boundary prior saliency map as the center of salient object to compute spatial saliency.Simulation result demonstrates that this methor can uniformly highlight saliency object and effectively suppress the background in natural scenes compared with Itti algorithm,contrast based method,graph based method,etc.