基于显著性和区域比较的自动物体提取算法

摘要

Facing the massive image data in internet era,how to automatically extract the object from images becomes a hot issue. Therefore this paper proposes a novel algorithm of automatic target extraction,it combines the superpixel,salience and region contrast.First, the algorithm makes superpixel segmentation on images to get several sub-regions.Then,it uses saliency detection to determine the initial target region.At last,based on sub-regions and initial regions,it combines spatial information and colour feature and uses region comparison method to segment the final target object.Result of the contrast experiment shows that this method can extract the object effectively and has certain robustness.%面对互联网时代海量的图像数据，如何自动地提取物体成为一个热点问题，为此提出一种结合超像素、显著性和区域比较的自动目标提取算法。算法首先对图像进行超像素分割，得到若干子区域；其次采用显著性检测确定出目标的初始区域；最后在子区域和初始区域的基础上，结合空间信息和颜色特征，利用区域比较法分割出最终的目标物体。对比实验结果表明，该算法能够有效地提取出目标，具有一定的鲁棒性。