The Enhancement And Denoising Of Palmprint Image Based On Partial Differential Equations

The palmprint recognition and identification becomes to be a newly developed biometric identification technology, since palmprint presents lots of unique advantages them other biological characteristics. The palmprint is identified through statistical model generally, and the process of palmprint identification follows the steps: image capture, pre-processing, feature extraction and choice, classification and decision making. It's crucial to extract the texture of palmprint image precisely. The captured images should be denoised and the feature should be extracted before matching.The database of palmprint image is very large because the image usually covers a large area and has numerous texture, and each texture has its own length, direction and width, the texture intersects in an irregularly way. All these make the image has a low contrast and easy to be influenced by noises. The image is a kind of texture image which has its own periodic texture, and the detail information of the image is the foundation for identification. The method of low-pass filter can denoise part of noises in the image, but the contrast between ridges and valleys in palmprint image is decreased simultaneously. The method based on partial differential equations(PDEs) turn out to be a good choice for denoising image,which can smooth part of an image in a great scale and the edge in a small scale by choosing diffusion coefficient suitably,and make sure that the diffusion follows the direction of edge, so the method can solve the problem whether to keep the edge or denoise it. Perona and Malik developed a non-linear anisotropic diffusion model based on the equation of thermal transmission which opens a new field for research and the application of image processing. A lot of researchers improved the P&M model by trying to construct more effective smooth filter with edge protecting and obtained lots of achievements.The main efforts of this paper are the denoising and enhancement of the plamprint image base on PDE model, and producing a new diffusion coefficient to improve the P&M model according to requirements of the characteristic of palmprint image and the demands of pre-processing the image. The diffusing speed is determined by the grade and second derivative which is imported by the diffusion coefficient function. The improved model has better diffusing characteristic, and performs the coherent diffusion based image enhancement in structure information preservation.The theory and the outcomes of the experiment indicate that the improved P&M model can success the P&M model's advantages, such as denoising in the smoother part of the image, keeping the details of the image at edge, and keeping the information of singular point singularity of the image, so the improved model can export an perfect pre-process image. The requirements of image processing in two aspects are met by using the improved model, the first one is noise removal and the second one is to enhance the characteristic enhancement of the image.It's also the first application of a method based on PDEs in pre-processing of palmprint image.