| DC 欄位 |
值 |
語言 |
| DC.contributor | 電機工程學系 | zh_TW |
| DC.creator | 范文長 | zh_TW |
| DC.creator | PHAM VAN TRUONG | en_US |
| dc.date.accessioned | 2013-6-14T07:39:07Z | |
| dc.date.available | 2013-6-14T07:39:07Z | |
| dc.date.issued | 2013 | |
| dc.identifier.uri | http://ir.lib.ncu.edu.tw:88/thesis/view_etd.asp?URN=965401601 | |
| dc.contributor.department | 電機工程學系 | zh_TW |
| DC.description | 國立中央大學 | zh_TW |
| DC.description | National Central University | en_US |
| dc.description.abstract | 影像切割是電腦視覺與物件辨識技術中重要的分枝.影像切割可視為將影像分割成前景與背景的技術. 目前存在的影像切割技術,如直方圖閥值(histogram thresholding), 區域成長分割法(region-growing), 群集法(clustering)等. 在這些分割技術中，變分方法(variational methods)已被證明是一個可行的影像切割方法。這些方法具有完善數學模型的優點。在影像切割中，透過求解能量泛函問題，變分方法可以萃取物體的邊。
本論文旨在開發基於變分方法的影像切割模型。在這篇論文中的第一個模型，我們提出了一種無監督的主動輪廓的方法進行異構圖像分割。在模型中，我們定義了能量泛函，其中包括區域和全域的能量。本論文獨特地利用影像強度在每一個局部影像建立區域能量, 同時, 根據輪廓內外圈的影像強度機率分佈之亮度稠密度距離(density distance)計算全域能量. 區域能量適合處理強度不均勻之影像. 與此同時, 全域能量有助處理降低雜訊,初始輪廓位置和輪廓蔓延等問題. 本論文提出之數學模型以應用於各種不同形式醫學影像之影像分割問題. 如磁振造影(MRI)，超音波(ultrasound)，電腦斷層(CT)圖像以及在醫療影像之三維器官模型等.
在本論文中提出的第二個模型解決影像中的不均勻性與快速收斂能量泛函最小值等問題. 首先, 我們整合內外圈每一個畫素的模糊隸屬函數(fuzzy membership function)
程度建立能量泛函.我們利用直接計算能量變化取代傳統梯度下降法計算能量泛函最小值問題. 本論文提出之數學模型有助於影像分割的快速收斂以及模糊邊界與強度不均勻性. | zh_TW |
| dc.description.abstract | Image segmentation is an important branch of computer vision and object recognition. It can be considered as the technique of separating an image into semantic parts including foreground and background. There have been many methods for image segmentation such as histogram thresholding, region-growing, clustering, and so forth. Among these techniques, variational methods have been shown to be a promising approach for performing the segmentation task. Variational methods have many advantages since they benefit from well-established mathematical models. In the context of image segmentation, variational methods allow extracting object boundaries through solving optimization problems for energy functionals associated with the image to be segmented.
This dissertation aims at developing models for image segmentation based on variational methods. In the first model in the thesis we propose an unsupervised active contour approach for heterogeneous image segmentation. In the model, we introduce an energy functional which includes both local and global energies. In particular, we construct the local energy using intensity in each local region of image. We concurrently establish the global energy based on the density distance between intensity probability distribution functions inside and outside the contour. The local energy allows dealing with images with intensity inhomogeneity. Meanwhile, the global energy helps alleviate problems associated with noise, initial contour position, and contour spilling. The model is then applied to segment different medical image modalities, such as MRI, Ultrasound, CT images, as well as 3D organs in a medical data set.
The second model presented in this dissertation involves addressing inhomogeneity in images and achieving fast convergence in minimizing the energy functional. In this proposed model, the degree of fuzzy membership function of each pixel to inside (or outside) contour is integrated into the energy functional. Especially, to minimize the energy functional, we directly calculate the alteration of the energy, instead of using gradient descent approach as in classical approaches. The proposed model therefore fast converges to the segmentation solution. In addition, the model can facilitate the segmentation task even in the presence of blurred boundary, and intensity inhomogeneity.
In sum, this dissertation presents two level set-based active contour models to perform image segmentation tasks. The proposed models utilize both global and local information to guide the motion of contour which helps the models effectively dealing with inhomogeneous images. | en_US |
| DC.subject | 主動輪廓 | zh_TW |
| DC.subject | 影像切割 | zh_TW |
| DC.subject | 整合局部與全域資訊 | zh_TW |
| DC.subject | 模糊能量 | zh_TW |
| DC.subject | 醫學影像切割 | zh_TW |
| DC.subject | Active Contours | en_US |
| DC.subject | Image Segmentation | en_US |
| DC.subject | Global and Local level set | en_US |
| DC.subject | Fuzzy Energy | en_US |
| DC.subject | Medical Image Segmentation | en_US |
| DC.title | 整合局部與全域資訊的等高線之主動輪廓於影像切割應用 | zh_TW |
| dc.language.iso | zh-TW | zh-TW |
| DC.title | Local and Global Level Set based Active Contours with Applications to Image Segmentation | en_US |
| DC.type | 博碩士論文 | zh_TW |
| DC.type | thesis | en_US |
| DC.publisher | National Central University | en_US |