无监督智能化图像分割与合成方法  

Unsupervised Intelligent Image Segmentation and Composite Method

作　　者：何春辉 HE Chunhui(Science and Technology on Information Systems Engineering Laboratory,National University of Defense Technology,Changsha 410073,China;Hunan Aike Human Resources Service Co.,Ltd.,Changsha 410208,China)

机构地区：[1]国防科技大学信息系统工程重点实验室,湖南长沙410073 [2]湖南艾珂人力资源服务有限公司,湖南长沙410208

出　　处：《智能物联技术》2023年第6期9-15,共7页Technology of Io T& AI

摘　　要：GrabCut图像分割算法在计算机视觉领域应用广泛,但它的短板是需要依赖人机交互进行区域选取,因此无法达到大规模图像智能化处理的要求。为消除人机交互操作和实现智能区域选取,本文提出了一种ROI智能区域选取方法以改进GrabCut模型,从而实现无监督的智能化前景图像分割,并结合像素替换方法完成前景图像和指定背景图像的无缝合成。实验结果表明,本方法可以有效实现通用领域前景图像分割与背景图像的合成任务。GrabCut image segmentation algorithm is widely used in the field of computer vision.However,its shortcoming is that it needs to rely on human-computer interaction to select regions.Therefore,it cannot meet the requirements of large-scale image intelligent processing.In order to eliminate human-computer interaction and realize smart region selection,this paper proposes an ROI smart region selection method to improve the GrabCut model,so as to realize unsupervised intelligent foreground image segmentation.Finally,the pixel replacement method is combined to complete the seamless composite of the foreground image and the specified background image.Experimental results show that this method can effectively realize the task of foreground image segmentation and background image composite in general domains.

关 键 词：图像分割 图像合成 GRABCUT Opencv-Python 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TP311[自动化与计算机技术—计算机科学与技术]