Hand Image Verification Method Based on PCA Eigenvectors

Abstract

In this research, the Principal Component Analysis (PCA) based method (namely Eigenhand), is used to verify persons from their hand’s image. Our approach treats the hand recognition/verification problem as an essentially 2D-problem rather than requiring recovery of 3D geometry, taking advantage of the fact that hand’s images can be described by a small set of 2D characteristics features. Here, the features are referred as “Eigenhands” because they represent the eigenvectors of the set of the trained and tested hands. The verification operation between the trained hand’s images (i.e. preserved in the Database) and the input “unknown” hand image is performed by utilizing the Minimum-Mean-Distance “MMD” criterion. Several amounts of different noises (i.e. Gaussian, Uniform, Salt-and-pepper) have been added to the test hand to measure the reliability of our presented verification system in the presence of the noise. Invariant tests against image rotation, resizing, shifting have also been carried.