Palmprint Recognition System With Preprocessing Based On Partial Differential Equations

In this dissertation, we establish a new palmprint recognition system based on PDE denoising and enhancement, where the nonlinear PDE models are taken as PM model and Ding's improved PM model. Afterward, we carry out a recognition experiment by using of PolyU palmprint database and analyze the recognition results. By comparing with the system using the same database (extracting subgraph by the largest inscribed circle, denoising by morphological wavelet), we obtain a better recognition result. The comparison between the recognition systems based on PM model and the improved PM model respectively, it shows that the improved PM model is a better method for denoising palmprint image.The dissertation is constructed by the following parts1. Location and extraction of subgraph: to carry out a pretreatment on palmprint image by mathematical morphology method, this can smooth the edge curve of palm; to tracing the profile of palm by 8 point neighbor algorithm; to produce a new algorithm according to the characteristics of the palmprint in PolyU to detect the corner and establish the coordinates; to rotate and locate the palmprint image, extract out the subgraphs.2. Denoising and enhancement of subgraph: to adjust the contrast by histogram equalization; to denoise and enhance the subgraph by PM and improved PM model.3. Identification of the palmprint: we use K-L transformation to identify the palmprint, where the distance is taken as 1 norm by a series experiments.4. Comparison and analysis of the result.