透過您的圖書館登入
IP:3.145.70.60
  • 學位論文

多重物體凸形輪廓定位之研究

THE RESEARCH OF MULTI CONVEX HULL BY LOCALIZATION

指導教授 : 陳永盛
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


基於人類視覺而言,若想從一幅影像中區分出各個獨立的物件,簡直是一件易如反掌的事。但直至今日,科學家們尚未發展出一套完全模擬人類視覺的系統。不過令人欣慰的是如今電腦科技的突飛猛進,也使得各方面的數位影像處理技術跟隨著提升,一些過去不易處理的問題也都可被輕鬆地解決。 然而在電腦視覺領域中不論是圖形辨識(Pattern Recognition)、影像復原(Image Restoration)與影像認知(Image Understanding)技術研究皆有物件輪廓萃取的需求,而且關於此一方面的學術發表、技術運用的文章也如雨後春筍般地出現。影像分割(Image Segmentation)的目標簡而言之便是各取所需,一般將分離後的資訊稱之為前景影像與背景影像。

並列摘要


The object contour is extracted from an image by human vision that is a piece of cake action. Scientist has not yet found a successfully system to explain and match human vision in present-day world. Whatever, the computational technique has lot of progress today that many problems of image processing can be solved more easily. In computer vision, the Pattern Recognition, Image Restoration, and Image Understanding are required the technique of object feature extraction. Moreover, the articles and applications about the object contour segmentation are announced as mushroom. The target of image segmentation is take whatever it need to application, usually application is separated an image as background and foreground. Therefor, we performed a method of multi-object segmentation based on the principle criterion of active contour models and dynamic programming in this thesis. There are many basic techniques have employed to realize the function of multi-object segmentation, they are histogram calculation, gradient distribution, thresholding, labeling, convex hull and etc.. The gray plane and gradient plane were composed these basic techniques into an iteration procedure to extract the object contours with convexity. The experimental results show our proposal can be implemented the aim of this thesis.

參考文獻


多重物體凸形輪廓定位之研究

延伸閱讀