以影像處理為基礎的影片中人臉置換系統

Abstract

在本論文中,我們提出一個完整的系統來實作影片中的人臉置換功 能。我們的系統會接受兩個輸入影片,主要目的是把其中一個影片裡 角色的人臉置換成另一個影片中的人臉。我們的系統分成三個主要的 步驟,首先我們會找出影片中的人臉位置,而且在找到的人臉上標出 主要的臉部特徵,其次是分析影片中人臉的表情和角度方位並且選出 最適當的替換表情,最後我們把選到的替換表情貼上並且把顏色做適 當的混成。我們的系統不需要任何3D的模型資訊,是全自動並且可以 產生逼真的效果。

In this paper, we present a complete system for plausible face replacement of videos. In our system, we take two videos as inputs. And we aim to replace the target human subject’s face with another human subject’s in the other video. Our replacement algorithm has three main stages. First, given an input video, we detect all faces that are present, and align such detected faces. Second, we analyze facial expressions of each detected faces and select candidate face images from source video that are most similar to the target face in pose and expression. Third, we blend candidate replacements to target video. Our approach requires no 3D model, is fully automatic and generates plausible results.