Abstract
Iris recognition is a well studied area in computer vision and pat- tern recognition. However most of the works are based in Gabor wavelets for feature extraction and either similarity matching or using a SVM. Even though this methods achieve a good performance in this paper we try applying HOG descriptors for the feature extraction to generate a highly dimensional space which is latter reduced using a random forest and finally compare different types of learning methods.
Descripció
Projecte realitzat mitjançant programa de mobilitat. ZHEJIANG UNIVERSITY, HANGZHOU. INSTITUTE OF ARTIFICIAL INTELLIGENCE.