基于交互式目标分割算法的影视后期抠像方法研究  

Research on post production image cutting method for film and television based on interactive object segmentation algorithm

作　　者：衣傲尘 尤俊杰[1] YI Aochen;YOU Junjie(Modern College of Northwest University,Xi’an 710130,China)

机构地区：[1]西北大学现代学院,陕西西安710130

出　　处：《电子设计工程》2024年第11期188-191,195,共5页Electronic Design Engineering

摘　　要：快速准确地提取出视频中的目标物体是影视后期抠像的关键环节,因此研究基于交互式目标分割算法的影视后期抠像方法。采用优化后的三帧间差分法检测影视后期视频图像内的感兴趣区域(ROI),以CV模型为基础,结合形状先验引导进化思想构建交互式目标分割模型,将待分割图像的先验形状信息与图像内实际目标轮廓信息作为交互信息,在交互式目标分割模型内引入面积激励项,由此在影视后期视频图像内ROI区域内分割出目标。将分割出的图像目标与待替换背景视频图像合成,完成影视后期抠像。实验结果显示,该方法能够有效确定视频图像中的ROI区域,在模型迭代30次条件下即可获取高精度的目标分割结果,抠像过程中的图像轮廓收缩率均控制在0.45以下。Fast and accurate extraction of target objects in video is the key link of post production image cutting for film and teleriso,so this paper studies the post production image cutting for film and teleriso method based on interactive object segmentation algorithm.The optimized three⁃frame difference method is used to detect the ROI in the post⁃film video images.Based on the CV model,the interactive target segmentation model is constructed with the idea of shape priori guided evolution.The prior shape information of the image to be segmented and the actual target contour information in the image are taken as interactive information,and the area incentive term is introduced into the interactive target segmentation model.Therefore,the target is segmented within the ROI region in the post⁃film video image.The segmented image target is combined with the background video image to be replaced to complete the image matting in the later stage of film and television.Experimental results show that this method can effectively determine the ROI region in video images,and can obtain high⁃precision target segmentation results under the condition of 30 iterations of the model,and the image contour shrinkage rate is controlled below 0.45 in the process of image keying.

关 键 词：交互式 目标分割 影视后期抠像 感兴趣区域 形状先验引导 激励项 

分 类 号：TN98[电子电信—信息与通信工程]