Study On Palmprint Recognition Based On Multiresolution Analysis

The multiresolution and multidirection analysis is a very important topic in digital image processing. It is a very effective way to analyze complex texture by transforming the image to different scales with different directions. Wavelet is a practical tool to implementation above mentions. Based on the advantages of wavelet, many hyper-wavelet algorithms are developed recently, among which Contourlets is the most effective one. Contourlets is not only able to analyze the image in different scales, but also be able to analyze in almost any directions.This paper applies Contourlets to the analysis and retrieval of palmprint images. Palmprint recognition is a very important topic in the biometric technology. The human palmprint is more stable, more clearly represented and easier to collect compared with other human biological characteristics, thus many palmprint recognition algorithms are developed. Because the palmprint images display the obvious anisotropy and the clear levels of texture, many of the algorithms are based on the multiresolution approach. Because the result of the non-subsampled Contourlets is more stable to the image change of rotation and shift, which are caused by the collecting stage, it is suitable to adopt the non sub-sampled Contourlets to analyze the palmprint texture. The main work of this paper is as follow:1. Introduces the basic theory of Contourlets and its application in digital image processing.2. Introduces the necessity of the pre-processing of palmprint images, and proposes the pre-processing methods to reduce the effects of rotation and shift which influence the retrieval, and fetches the key area of palmprint. The pre-processing method includes the thresholding, erosion operation, etc.3. Calculates the eigenvectors of the palmprint images using Contourlets coefficients and implements a palmprint images retrieval system.Experiments show that the methods are effective and efficient.