Efficient Sampling Strategy and Refinement Strategy forRandomized Circle Detection

• Main Authors: Hsi-ming Shen, 申希明
• Other Authors: Kuo-Liang Chung
• Format: Others
• Language: en_US
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/56576919340315401116

碩士 === 國立臺灣科技大學 === 資訊工程系 === 98 === Circle-detection is an important and fundamental operation in image processing and pattern recognition. This thesis first presents a new gradient line-based sampling strategy to determine a candidate circle and the determined candidate circle has higher probability to be promoted to a true circle when compared with the traditional randomized strategy; it leads to significantly computational effect. Further, for enhancing the detection accuracy, a new revoting-based refinement strategy is presented. Experimental results demonstrated that our proposed gradient line-based sampling strategy and revoting-based refinement strategy can significantly improve computing time performance and the detection accuracy for circle detection when compared with previous randomized related algorithms.