一种基于区域生长的图像显著区域提取算法

An Image Salient Regions Extraction Algorithm Based on Region Growing

提出了一种改进的基于区域生长的彩色图像显著区域提取算法。该算法在颜色梯度图上应用分水岭分割算法,改进了图像初始分割的效果。综合考虑分割区域的显著度和边界强度,选取恰当的种子区域。在区域生长时考虑分割区域间低层特征的相似性、区域是否位于物体边界以及区域间的可合并度,设计了三个区域可生长度评价函数。实验结果表明,该算法简单、快速,且能提取出更符合人眼视觉特征的显著区域。

An improved color image salient region extraction algorithm based on region growing is proposed in this paper.The algorithm improves the performance of initial segmentation by utilizing Watershed on color gradient image.Take salient maps and boundary strength based on the regions into account,proper seeded regions are selected.In the process of region growing,three function to evaluate the capacity of region growing is revised,aiming at measuring the similarity of low features of adjacent regions,the relative position between regions and subjects,and the degree of compact connection.The experimental results dem onstrate that this algorithm is simple,fast and is able to extract the salient regions in line with the characteristics of the human visu al system.