Palmprint recognition using Gabor magnitude code

Because of its robustness, user friendliness, low cost and high accuracy, palmprint recognition has been widely studied in the past ten years. Various feature extraction and matching schemes have been proposed, among which the Gabor phase and orientation codes are very effective and efficient for palmprint representation and matching. Although these methods are adopted in the online palmprint recognition systems, they neglect the Gabor magnitude information in coding. On the other hand, existing Gabor magnitude based methods could not be combined with the well developed Gabor phase and orientation codes efficiently because they use different feature extraction and matching procedures. In this paper, a novel Gabor magnitude feature extraction algorithm is proposed. The algorithm represents Gabor magnitude information by binary code which is obtained by adaptively thresholding the image. The proposed magnitude code could be readily combined with the Gabor phase and orientation codes. Experimental results on a large public palmprint database show that the accuracy could be improved by fusing the proposed Gabor magnitude features with original phase or orientation features.