Selective Algorithm Outline (SAO); An Alternative Approach for Fusing Different Palm-Print Recognition Algorithms

摘要

In recent years, there has been renewed interest in integrating holistic-based and feature-based palm-print recognition approaches in a cascade way for improving the recognition performance. They use a holistic-based approach as a filter to narrow down the database for further feature-based matching. Although cascaded fusion schemes have achieved good performance, they suffer from two shortcomings. First, if the training samples of true class are not in the narrowed gallery set, feature-based approach fails to identify the correct class. Second, the feature-based approach is unnecessary if the holistic method is able to identify the correct class solely. In this paper, we present an efficient palm-print recognition algorithm based on cascaded fusion of holistic-based and feature-based algorithms, where best algorithm is specially selected for classifying individual testing samples. It uses a holistic-based approach for classifying testing samples and alternates it with a feature-based approach just in case the classification result of holistic-based algorithm is not sufficiently confident. The experimental results on Poly-U, CASIA and UST databases show that the proposed method has a noticeable performance improvement compared to baseline methods.