基于信息弥散机制的图像显著性区域提取算法

An image saliency object detection algorithm based on information diffusion

为了更好提取图像的显著性区域,提出基于信息弥散机制的图像显著性区域检测算法。在所提算法中,首先将图像分割成超像素,根据图像中显著性区域频率变化比较大的特性,生成图像显著性区域的高频节点;然后针对高频节点利用凸包运算寻找显著性区域的种子节点,最后使用二阶高斯-马尔科夫随机场信息弥散方法在图像中对种子节点进行显著性区域信息扩散,得到图像的显著性区域。试验结果表明,利用二次规划求解每个数据之间的线性关系进行信息扩散,能够达到避免阈值选择和信息精准分类的效果,其结果优于同类的图像显著性区域检测算法。

In order to better extract salient regions in images,we proposed an image salient region detection algorithm based on information diffusion mechanism. The proposed algorithm was divided into three steps. First,we segmented an input image into superpixels which were represented as the nodes in a graph. The node with high frequency was generated by the characteristics of the salient regions. Then,according to high-frequency nodes,convex hull computation was used to generate the saliency seeds of the salient object area. Finally,based on the seeds obtained by convex hull computation,the second-order Gaussian-Markov random fields were used to diffuse the information from saliency seeds to others,thereby forming the saliency region for a given image. The experimental results showed that the quadratic programming solution exploited to compute the weights between the nodes can effectively avoid threshold selection and enhance robustness accordingly. In addition,the proposed method performed better than the other state-of-the-art methods.