利用影像曲面之曲率資料重建影像曲面

Abstract

在這篇論文當中，我們提出一個新的影像重建方法：利用影像曲面中曲率 大的特徵點來重建影像。這些特徵點可以很容易掌握，並方便進一步的影 像處理。我們先重建影像截面，分析重建的品質和穩定性，然後再提出幾 個影像內插的方法，使重建的結果更平滑。為了簡化特徵點的表示方法以 促進影像壓縮，我們也討論了特徵點的連結。由簡單的實驗當中，我們可 以預見，此種重建方法將可以應用在高倍率的影像壓縮上。 In this thesis, we present a novel approach which reconstructs an image by using high-curvature points of the image surface. These feature points can easily be handled and used for different applications in image processings. We reconstruct the image profile first and analyze the quality and stability of the reconstructed results.Then, several interpolation methods are proposed to improve the smoothness of the reconstructed image. To simplify the representation and facilitate further compression, the linking of feature points is also discussed. A simple but promising experiment is then demonstrated to show the capability of doing image compression for very-low- bit-rate applications.