電腦視覺系統使用攝影鏡頭做數位影像之擷取,此擷取的影像經前置處理改善影像視覺效果,以增加後續影像處理的成功率。後續作業以型態學影像分析和物體特徵的辨識,來分離物體與背景已獲取我們需要的資訊。本文所探討的主題在PC 的平台上,使用Matlab 強大的矩陣運算和影像處理的能力,計算那些複雜的數學式,探討不同演算法在處理影像辨識上的研究。 本文目的是在利用現有的七種演算法 : (1)整體臨界值法(Global image thresholding) (2)局部適應臨界值法(Local adaptive thresholding) (3)自動臨界值法 (Auto thresholding-isodata) (4)最佳臨界值法(Optimal thresholding)(5)模糊C-means分群法 (Fuzzy C-mean clustering ) (6)疊代法(Iterative Method) (7) Global and Region Features 來求取臨界值。本文運用不同的演算法來簡化圖形,來取得二值化圖像,分離影像的前景與背景並分析實際運算上所會面臨的辨識正確性與效率問題,再以型態學影像處理方式做後續處理,進行影像二值化後的相關運用。
Computer Vision Systems acquire digital images by means of camera lens. The quality of these images are greatly improved through pre-process to the possibility of success of follow up process. The objects and backgrounds can be separated in order to catch the information needed during the analyzing process of morphology and object characteristics identification. The Matlab platform contains strong abilities of matrix calculation and image processing; hence, by means this powerful tool, this article aims to compute the complex mathematics for the purpose to discuss the research of different algorithms on digital image processing. This article utilizes the seven algorithms : (1) Global image thresholding (2) Local adaptive thresholding (3) Auto thresholding - isodata (4) Optimal thresholding(5) Fuzzy C-mean clustering (6) Iterative Method (7) Global and Region Features in search of the proper threshold value. This article applies various algorithms to simplify the images in this research to obtain binary images, during which accurate and efficiency will be analyzed as well. The follow-up procedure is worked in use of morphological processing for post operation out of further related applications after binary images.