数字抠像是图像处理、视频编辑和电影制作中的关键技术.通过数字抠像,从图像或视频的背景中精确地分离出前景, 是计算机视觉领域的重要问题.本文首先介绍了目前数字抠像的交互方式,然后把抠像技术分为基于颜色采样、 基于像素相似性、基于能量函数以及基于机器学习的四类技术,介绍和分析了其中的典型算法和最新研究成果, 并对这些算法的测试结果进行了定量和定性比较,最后总结了数字抠像技术目前的研究状况和未来的发展方向.
Digital matting is the key technology in image processing, video editing and film-making applications. It refers to the problem of extracting the foreground objects in the images or videos accurately, which is an important issue in the field of computer vision. In this paper, the interaction modes for digital matting are introduced firstly. After that, matting techniques are divided into four categories: color sampling based, pixel affinity based, energy function based and machine learning based. Classic algorithms and the latest research progress are presented and analyzed, followed by quantitative and qualitative evaluations of these techniques. Finally, the research progress is summarized and future research directions are suggested.