Extraction Of Palmprint In Multiscaler Base On Wavelet Transform

In Traditional Chinese Medicine(TCM), palmprint diagnosis is an important diagnosis method with a long excellent history and good diagnostic results. The objectivation and automation of palmprint diagnosis is of great significancy for TCM and will improve the effect of machine diagnosis in TCM. The line features of palmprint play a crucial role in palmprint diagnosis. Consequently, a high-accuracy method for extraction of line features of palmprint is a primary task of automatic palmprint diagnosis.However, due to the characteristics of heavy noise and low contrast of a palmprint image, it is a challenge in automatic palmprint diagnosis to extract various line features of palmprint as correctly as possible.The multiresolution wavelet transform analysis is a powerful technique in signal analysis, which has an excellent resolution capacity on the space-frequency plane, and has been widely applied in image processing.Firstly, with analysis on the characteristics of palms'profile from a palmprint image database, an algorithm for image alignment and segmentation is proposed, which is of great benefit to the robust, the de-noising and the rotation and scaling invariance of the palmprint line extraction algorithm described later.Secondly, an improved module maximum of binary wavelet transform algorithm is presented with introducing multiresolution analysis, and a B-spline wavelet is constructed to extract the ridge of palmprint,. Then a novel line extraction algorithm based on wavelet filter is proposed. The experimental results of the two algorithms show that both of them are good at noise suppression and brightness elimination and can effectively extract various palmprints from heavy noise and low contrast palmprint images.In addition, the mathematical foundation of wavelet analysis theory and its excellence are analyzed with a brief description of multiresolution analysis and the Mallat algorithm and their extension to 2-D signal analysis. And then the theoretics of biorthogonal wavelet which was variously used in image processing is introduced.And several edge detection algorithms and their characteristics are also discussed in this thesis.