連續虹膜影像中利用影像品質評估篩選適用於辨識之影像之研究

Abstract

在虹膜影像辨識中，輸入之虹膜影像好壞將會影響到後續之虹膜辨識之結果，若同一受測者可於同一時間拍攝多張虹膜影像，並於多張虹膜影像中選擇較佳之影像進行辨識，將可提升虹膜辨識成功率與效率。但目前之虹膜影像評估方法並不能將不適合辨識之虹膜影像挑選出來，且其分類為低品質之影像並不一定不適合做虹膜辨識。針對上述之問題，本論文設計一套虹膜影像優先序系統，主要目的為可於虹膜辨識前進行虹膜影像篩選，可將不適合辨識之虹膜影像先行排除掉，並依據影像品質評估之結果將可辨識影像進行排序，減少虹膜辨識對於不適合辨識之影像的處理時間。 在本論文的研究中，本論文的貢獻有以下二點：第一，本論文提出一新的品質評估方法來判斷輸入之虹膜影像其虹膜與瞳孔是否定位錯誤，使得可預先排除掉定位錯誤之影像，達到虹膜辨識前之快篩處理；第二，設計一個優先序機制系統，可以依據虹膜品質評估之結果將影像依序排列，使得可於同一受測者之多張影像中挑出最適合的影像進行虹膜辨識，以增加虹膜辨識之效率。

Iris image quality critically determines iris recognition performance. If the subjects can be taken multiple iris images and select the best one for recognition by the system simultaneously, the recognition rate and efficiency of iris recognition will be improved. However, the existing iris image quality assessment methods cannot pick out the improper images. In this thesis, we propose an iris image priority order system, which can exclude the improper images, sort the identifiable images, and reduce the execution time of iris recognition. There are two contributions of this thesis. First, this thesis proposes a new assessment method to evaluate whether the iris localization is correct, which can exclude the image of position error before iris recognition. Second, this thesis proposes an iris image priority order system, which can sort the images by the result of the quality assessment system. Therefore, the most suitable iris image can be picked out from multiple images of the same person to increase the efficiency of iris recognition.