繪圖處理器於數位全像顯微術之研究

Abstract

在本論文中，主要探討繪圖處理器的平行運算特性及其應用於離軸式數位全像顯微系統之數值重建與相位展開，以達到即時獲得量測樣本的三維輪廓。本研究工作首先探討繪圖處理器平行處理的特性及其與串行處理運算效能之比較，並且更進一步的利用繪圖處理器於數位全像術之數值重建運算，將其演算法平行化。在重建的過程中將結合濾波處理以提高重建影像對比度，最後將所量測到樣本的相位資訊經由相位展開演算法來獲得樣本的三維資訊，並且藉由繪圖處理器高速運算的能力，以達到即時運算與顯示的效果。本研究證實了使用繪圖處理器的高速平行計算特性於數位全像顯微術具有即時重建與相位展開的潛能。內文中將提出相關模擬與實驗結果，並加以分析與說明。

In this study, we investigate the properties of parallel computing with graphics processing unit (GPU) and its application on off-axis digital holographic microscopy (DHM). First, we study the properties of parallel computing with graphics processing unit and compare the performance to the serial computing with central processing unit (CPU). And the parallel computing with graphics processing unit can be used to integrate a filter algorithm into the numerical reconstruction calculation to get a good-quality reconstructed image and reduce the calculation time. The digital holographic microscopy equipped with the graphics processing unit can also accelerate the phase unwrapping procedure for obtaining the three-dimensional profile of a specimen in real time. This work demonstrates that high-speed parallel computing of graphic processing unit in digital holographic microscopy has great potential to perform numerical reconstruction and phase unwrapping in real time. The analytical and experimental results are presented and discussed.