浅析基于机器学习的威亚擦除技术

An Analysis of Wire Remove Technology Based on Machine Learning

威亚擦除(Wire Remove)是较为常见而重要的数字影视合成技术,数字合成师需要对拍摄特技镜头过程中所使用的威亚(Wire)进行擦除以避免画面的穿帮。本文将跨界融合数字影视合成技术、计算机视觉以及人工智能技术,以数字影视合成技术中的威亚擦除作为人工智能和数字影视技术跨界融合的研究对象。围绕目前较为成熟的“影片修复”“图像补全(Image Inpainting)”、Adobe软件中的Content-Aware Fill(内容识别填充)技术以及Nuke13中新增的CopyCat机器学习工具集等技术进行综合研究,为当下数字影视合成中的威亚擦除提供新的思路和解决方案。

Wire remove is a common and important technique in digital film and television compositing,where the digital compositor needs to erase the weave used during the filming of a stunt shot to avoid goofs.In this paper,we will integrate digital film and television composition technology,computer vision technology and artificial intelligence technology,and take the wire remove in digital film and television composition technology as the research object of the integration of artificial intelligence and digital film and television technology.The research will focus on current mature technologies,such as"film restoration","image inpainting","Adobe Content-Aware Fill",as well as CopyCat,the new machine learning toolsets in Nuke13,to provide new ideas and solutions for the current digital film and video compositing in the wire remove.