多視點視訊產生、傳輸與分析-子計畫一：戶外場景之多視點視訊產生技術研究

Abstract

本計畫為“多視點視訊產生、傳輸與分析”群體計畫之子計畫一。近年 來，立體電影與立體電視已蔚為風潮，隨著電腦硬體設備的快速進步及 取像設備的日漸普及，多年前就被提出來的多視點視訊想法已逐漸成為 一個可行的做法。只是，目前的多視點視訊取像系統大多龐大笨重，並不 適用於戶外場景，在本計畫中，我們希望建構一套比較方便於攜帶的多視 點視訊取像系統，使用數量較少的攝影機架設在彼此相距較遠的位置， 這樣的系統所產出的視訊品質雖然會遜於前述的龐大取像系統，但卻可 以讓戶外場景的取像過程變得方便彈性許多。 我們將以三年的時間進行研究討論。第一年度將著重於雛型系統的建 立、校正技術的建立及spectral matting 技術的開發；第二年度將著重於三 度空間相對深度的估測與影像比對技術的研究；第三年度則將著重於透過 ray space 空間將多攝影機視訊彙整成多視點視訊資料。

This project is a sub-project of the joint-project “Multiview Video Generation, Transmission and Analysis”. In recent years, 3D movies and 3D TV’s are getting popular. Besides, with the advancement of information technology and image/video capturing devices, applications related to multi-view videos are becoming feasible. Unfortunately, existing systems for the capturing and generation of multi-view videos are far too cumbersome for outdoor usage. Hence, in this subproject, we aim to develop a portable multi-view video generation system suitable for outdoor scenes. This is a three-year project. In the first year, we will focus on the installment of a prototype system, together with the development of the camera calibration methods and the spectral matting technique. In the second year, we will focus on the 3D depth estimation from a single camera and the development of image correspondence technique. In the third year, we will discuss the generation of multi-view videos based on the fusion of multiple video inputs in the ray space.