臺灣博碩士論文加值系統

本論文中,針對黑白影提出一個影像分割系統.這個系統採用監督式的架構
而分為兩個階段來完成:初始值設定階段與迭代階段.在階段一中針對影像
中以每一個相素為中心用一調適性視窗收集適當的資料,而視窗的大小是
由每一相素的週遭相素分佈情況決定,以 AIC 準則來衡量視窗中的影資料
是否適合分類成我們要的類別數. AIC 決定視窗大小後,以K-類平均分類
法得到相素的初始值及影像模式中所需的資料.在階段二中以階段一中找
到大區域特性加上馬可夫隨機場以考量空間領域束縛特性,組合成一影像
模型,再加以分割.本系統調適化的分類法使處理過的影像不至於再發生失
去影像細節重點,為此本文也將和前輩的做法做一比較以說明本系統的良
好行為.

In this thesis, we present an algorithm for segmentating gray
images. This algorithm accounts for local intensityations
includes spatial constraints in the image .Local intensity
variations are accounted for local average estimation ovre a
sliding window whose size determined by minimizing acriterion
based on AIC information criterion. We use K_means algorithm to
estimate the initial class and local intensity variations in
the capture window. Spatial constraints are included by the use
of a Markov random field model. This segmentation algorithm is
adaptive. The segmented images are preserved improtant details.
Moreover this algorithm is suitable for parallel computing. It
is easy to implement in hardware.