In order to ensure the existing saliency region model on color image can be applied to stereo image, an extension framework for merging color-produced saliency map and depth-produced saliency map to obtain the significant results of stereo images is provided. Among them, the former uses existing color image saliency detection algorithm, the latter uses the algorithm based on the regional contrast image depth significantly detection which is proposed in this paper. In addition, a two-layer saliency region detection model based on stereo image is also presented, considering the color information, the depth information and the correlation from two aspects. The experimental result shows that the model could accurately position salient objects in the stereo image simply and efficiently, and improve the precision rate and recall rate of the detection algorithm.%为了确保现有的彩色图像的显著性模型可以应用于立体图像，提供一个扩展框架，融合彩色图像和深度图像产生的显著图最终得到立体图像显著性结果。其中，前者利用现有的彩色图像显著性检测算法，后者使用基于本文提出的基于区域对比度的深度图像显著性检测算法。此外，还提出一种基于立体图像显著性检测双层模型，从两个层面综合考虑了颜色信息、深度信息及其相关性。实验表明，本文的模型可以简单高效的在立体图像中精确定位显著性对象，并且提高了检测算法精确率和召回率。
展开▼
