模糊幾何在數位影相處理上的應用

Abstract

本篇論文以模糊幾何 (fuzzy geometry) 的方法，研究數位影像處理 (digital image processing)。Rosenfeld (1988)，Pal (1988, 1992)， 和 Ghosh (1992) 的演算法則 (algorithm) 係利用主體物 (object) 與 背景物 (background) 分離的技巧來作分割，而此演算法則在計算上是很 花費時間的。本文利用基因遺傳演算法則 (genetic algorithm) 來解決 同樣的問題，而本演算法則能有效的減少計算時間。同時，我們將演算法 擴充至多重臨界值 (multithresholding)，以 K-means 演算法則 (K- means algorithm) 及模糊幾何來處理影像。因此， 這些輸出影像並不僅 僅只是黑白二階 (bilevel) 的影像。 The thesis investigates the digital image processing problem by using fuzzy geometry. Algorithms provided by Rosenfeld (1988), Pal (1988, 1992), and Ghosh (1992) use the segmentation method for object/background classification, but the algorithms are expensive to compute. In this thesis, we propose algorithms using genetic algorithm to solve the same problem, and they are found to be successful to reduce the computing time. Meanwhile, we also make an extension of multithresholding by using K-means algorithm and fuzzy geometry to deal with the images, so that the output images are not just bilevel images.