Palmprint attracts increasing attention thanks to its several advantages. 1st-order textures have been widely used for palmprint recognition;unfortunately, high-order textures, although they are also discriminative, w...
详细信息
Palmprint attracts increasing attention thanks to its several advantages. 1st-order textures have been widely used for palmprint recognition;unfortunately, high-order textures, although they are also discriminative, were ignored in the existing works. 2nd-order textures are first employed for palmprint recognition in this paper. 1st-order textures are convolved with the filters to extract 2nd-order textures that can refine the texture information and improve the contrast of the feature map. Then 2nd-order textures are used to generate 2nd-order Texture Co-occurrence Code (2TCC). The sufficient experiments demonstrate that 2TCC yields satisfactory accuracy performance on four public databases, including contact, contactless and multi-spectral acquisition types. Moreover, in order to further improve the discrimination and robustness of 2TCC, we propose Multiple-order Texture Co-occurrence Code (MTCC), in which 1st-order Texture Co-occurrence Code (1TCC) and 2TCC are fused at score level. 1TCC is good at describing minor wrinkles;while 2TCC does well in describing principal textures. Thus the combination of both can describe the palmprint features more comprehensively. MTCC achieves remarkable accuracy performance when compared with the state-of-the-art methods on all public databases.
In the past few years, the need for accuracy and robustness against luminosity variations has drawn a considerable share of the palmprint research toward coding-based approaches. However, on the downside coding-based ...
详细信息
In the past few years, the need for accuracy and robustness against luminosity variations has drawn a considerable share of the palmprint research toward coding-based approaches. However, on the downside coding-based approaches require a high computational cost. On the contrary, while holistic-based palmprint recognition methods are easy to implement and have low computational burden, they usually do not result in a highly desirable accuracy. As a result, more recently hybridization of the holistic-based and coding-based methods has gained a boost. These hybridization schemes take advantages of both holistic and coding information to achieve a better performance. However, their computational burden due to incorporating the coding approach is still much heavier than the holistic methods. In this paper, we propose a new hybridization scheme based on Anisotropic Filter (AF) coding and the two-phase test sample representation (TPTSR) for the palmprint identification. In our scheme, the coding-based method is only applied on a super narrowed gallery in order to measure the classification confidence for a given test sample. Then, we apply our Guided Holistic (GH)-basedmethod for classifying the test sample if the holistic-based algorithm is not sufficiently confident. Experimental results demonstrate the efficiency of our method in enhancing both the complexity cost and the accuracy of the results.
暂无评论