本論文提出cDNA微陣列影像模糊濾波器之設計,使用於cDNA微陣列影像復原,以模糊推論系統設計雜訊偵測器,以不同濾波器之復原結果設計模糊推論規則,判斷是輸入像素向量否受雜訊污染,並以粒子族群最佳化雜訊偵測器,提高雜訊判斷效能。首先以影像預處理器將斑點前景與背景區域分離,以前景區域輸入所提出之雜訊偵測器,判斷輸入像素向量是否遭受汙染,基於選擇性濾波器,若判斷為乾淨像素向量,以輸入像素向量不變輸出;反之以濾波器復原輸出,並以斑點顏色比特性加強濾除結果。最後實驗證明本論文所提出之影像濾波器之復原效能勝過已存在之著名方法,且可應用於cDNA微陣列影像復原。
This thesis presents a fuzzy-based filter for removal of impulse noises for cDNA microarray image. The proposed filter is a kind of switching filters. It first employs a fuzzy inference system (FIS) to play a role for a noise detector which can detect whether a pixel vector of a cDNA Microarray image is corrupted. Meanwhile, it utilizes the particle swarm optimization (PSO) algorithm to optimize the noise detector so as to improve the accuracy of the noise detection. If a center pixel vector of a siding window is estimated as a noise-free pixel vector by the noise detector then the center pixel vector is unchanged to be the outputted pixel vector. Otherwise, the pixel vector is restored by the center-weighted median filter and enhanced within a proposed filter based on the color-ratio of spot. The experimental results demonstrate that proposed filter outperforms the existing well-known filters. Therefore, it can be applied in the cDNA microarray image restoration.