Abstract: In order to overcome the problems of nonuniform in salient object highlight and weak ability of anti-background existing in traditional approaches of saliency detection, a salient object detection method based on multi-scale contrast and Bayesian model is proposed. Firstly, the source image is segmented into a series of superpixels with same color feature, and multi-scale segmentation maps are calculated by K-means algorithm. Secondly, background priors and convex-hull center prior computing multi-scale saliency maps are adopted, and then a coarse saliency map is calculated through weighted summation. Finally, based on the rough region, a prior map is computed for the Bayesian model to achieve the final saliency map. Compared with 6 state-of the-art methods on publicly available datasets（MSRA-1000）, the simulation results demonstrate that the salient object detection approach proposed in this paper performs more uniform in highlight salient object,  with higher precision ratio and lower mean absolute error.

Key words: multi-scale, Bayesian model, background priors, salient object

摘要： 针对传统显著目标检测方法中目标不能均匀高亮，背景噪声难以抑制的问题，提出了一种融合多尺度对比与贝叶斯模型的显著目标检测方法。将图像分割为一系列紧凑且颜色相同的超像素，并通过K-means算法对所得超像素重聚类得到多尺度分割图；引入背景先验及凸包中心先验计算不同尺度下的显著图，并加权融合成粗略显著图；将粗略显著图二值化得到的区域假定为前景目标，再计算观测似然概率，使用贝叶斯模型进一步抑制图像的背景并凸出显著区域。在公开数据集MSRA-1000上与6种主流算法进行对比，实验表明提出的算法相比其他算法能更均匀地高亮显著目标，有更高的查准率和更低的平均绝对误差。

关键词: 多尺度, 贝叶斯模型, 背景先验, 显著目标