螢光眼底畵像의 特徵領域分割 및 認識 = Feature region segmentation and recognition of the ocular fundus fluorescein angiogram

다국어 초록 (Multilingual Abstract)

Image segmentation and recognition for the ocular fundus fluorescein angiogram with low contrast and unimodal histogram distribution are proposed.
For the segmentation, a locally adaptive difference image is generated and the feature points, which are used for the initial estimate for the region segmentation, are selected from the histogram of difference image. Then the label probabilities for the pixel classification are refined via the relaxation process.
The recognition procedure is separated into two steps: shape discrimination and description. In the shape discrimination process, component labeling is performed with the segmented feature regions. Using the shape factor defined by the area ration of the feature region to its circumscribed rectangle, all the feature regions are classified into the blood vessel and impaired retinal regions. Additionally the informations such as the position and the area of the feature regions are extracted. In the shape description process, the skeletal image of the classified feature regions are extracted using the distance transformed data. And inherent structure of the feature regions with the improved connectivity are obtained via the expanding process from the skeletons.
It is proved that the proposed algorithms are useful for the analysis of the fluorecein angiogram with the blurring and low contrast resulted from die leakage.

목차 (Table of Contents)

• 目次 = 1
• 緖論 = 2
• 螢光眼底畵像의 영역분할 = 8
• 1. 문턱값 분할방법 = 8
• 2. 螢光眼底畵像의 분할알고리즘 = 10
• 2.1 群分類의 基準點 선정 = 15
• 2.2 이온법을 이용한 화상분할 = 25
• 2.3 실험 = 30
• 特徵領域의 모양분석과 모양묘사 = 35
• 1. 분할된 영역의 식별과 표현 = 35
• 1.1 螢光眼底畵像의 특징영역 = 37
• 1.2 特徵領域의 分割 = 39
• 1.3 特徵領域의 識別 = 44
• 1.4 실험 = 49
• 2. 特徵領域의 描寫 = 52
• 2.1 距離變換 = 53
• 2.2 中心軸變換과 領域擴張 = 57
• 2.3 實驗 = 59
• 結論 = 63
• 參考文獻 = 65
• (Abstract) = 73