Local Kernel Mapping for Object Recognition

This paper proposes a new method, named Local Kernel Mapping (LKM), for object recognition. LKM is proposed to capture the nonlinear local relationship by using the kernel function. Different from traditional kernel methods for feature extraction, the proposed method does not need to reserve the training samples. To testify the effectiveness of LKM, we apply it on Local Binary Pattern (LBP), and the experiment results on palmprint show that LKM can improve the performance of the LBP method.