基於自動影像偵測與內容校正技術之大型畫像掃描系統

Abstract

圖書館或博物館內包含各式各樣的畫像，如大型畫像、油畫、水彩畫、細微畫像等等。為了建立一座完整的數位圖書館或博物館，需要發展出一套有效的系統來將這些畫像變成數位影像，以儲存到電腦裡。由於圖書館或博物館中存在的畫像一般多是比較大型的，而現在市面上所有的影像掃描器皆為小型且平面式的，無法用來掃描這些大型畫像，因此本論文研究設計了一套特別的大型掃描系統，以掃描這些畫像內容。首先，我們提出了一個新的自動影像偵測法，此方法可用來控制一具架於一X-Y二維移動平台上的攝影機，使其作全自動來回移動並拍攝大型畫像。另一方面，因影像分多次拍攝，影像之間有甚多差異或變形，包括位置、彩色、打光、幾何扭曲以及透視變形等諸多問題，在接合影像的過程中皆須加以克服，才能獲得一大張完美無暇的整合影像。在本篇論文裡，我們提出了一系列方便且有效的演算法來解決這些問題。良好的實驗結果證明了本系統的可行性。

A library or museum contains various types of pictures, such as large-scale picture, painting, miniature, etc. To establish a complete digital library or museum, it is necessary to develop an effective scheme to take digital images of pictures. Because the size of a picture in a library or museum is usually larger than the scanning area of a common scanner, the scanner cannot be used to scan the picture. For this reason, a special-purpose scanning system for scanning large-scale pictures is proposed in this study. First, a new approach to automatic image detection is proposed to control a camera mounted on an X-Y table automatically to scan a large-scale picture into several image pieces. Next, since the image pieces may be taken under various conditions of positioning, lighting, geometric distortion, perspective transformation, etc., a lot of problems arise in the process of integrating the image pieces to set up a complete image for the picture. In this study, we propose a set of convenient and efficient algorithms to solve these problems. Good experimental results show the feasibility of the proposed system.