题目眼前节光学相干层析图像的图像去噪与角膜厚度自动测量的新算法.PDF

題目：眼前節光學相干層析圖像的圖像去 噪與角膜厚度自動測量的新算法 Title ：Novel approaches to image denoising and corneal thickness automatically measurement in anterior segment optical coherence tomography images 姓 名 ： 杜文亮 Name 學 號 ： 1109853J-II20-0019 Student No. 學 院 ： 資訊科技學院 Faculty 課 程 ： 理學碩士 Program 專 業 ： 計算機與通訊技術 Maj or 指導老師 ： 田小林 Supervisor 日 期 ： 2014 年4月 Date Abstract Optical coherence tomography (OCT) for imaging biological tissue, which combines the advantages of optical heterodyne detection and scanning tomography, provides the ability of high sensitivity, high-axial resolution and the non-invasion. The whole system is compact and its low cost makes wide application in clinical medicine. This study focuses on the image denoising and corneal thickness measurement in anterior segment OCT (AS-OCT) images. There are two main challenges in image processing and corneal thickness measurement of the AS-OCT images. One is that the OCT images suffer from speckle noise, which re- duces image contrast and makes boundaries difficult to be distinguished from background. This study proposed two novel image denoising algorithms based on wavelet transform. One is about dynamic threshold segmentation (DTS), the other is about dynamic filtering (DT). The DTS removes the noise in low frequency sub-band. In contrast, DT identifies the region of interest in low frequency sub-band, then uses the sampling-based sub-band adaptive algorithm to suppress the speckle noise. Both of two new approach preserve more de