Palm-print Recognition Based on DCT Domain Statistical Features Extracted from Enhanced Image

In this paper, a feature extraction algorithm for palm-print recognition is proposed based on statistical features of two-dimensional discrete cosine transform (2D-DCT), which efficiently exploits the local spatial variations in a palm-print image. First, adaptive median filtering followed by Top-Hat transform is employed on a given palm-image to obtain palm-line enhancement by reducing the effect of noise and lighting variations. Unlike conventional median filtering, adaptive median filtering operates only on pixels, which are not structurally aligned and can preserve detail while performing overall smoothing operation. The entire enhanced image is segmented into several small spatial modules and 2D-DCT is performed on each module. Instead of considering all DCT coefficients, a set of statistical features are extracted in DCT domain, which drastically reduces the feature dimension and precisely captures the detail variations within the palm-print image. From our extensive experimentations on different palm-print databases, it is found that the performance of the proposed method in terms of recognition accuracy and computational complexity is superior to that of some of the recent methods.