Abstract: Finger vein recognition had attracted great attention for its unique advantages. For the finger vein images we obtained from the hardware system often contain severe noise, shadow and other issues, denoising of the low quality images had become the key to the recognition process.For this reason, a new image denoising method based sparse representation for finger vein image was proposed in this paper.Image denoising method based on image sparse decomposition is different from the traditional image denoising methods. In this method, image corrupted by noise is decomposed into two parts. One part is the image sparse components which are related to image information. Another part, which remains after the image sparse components are subtracted from the image, is regarded as noise. Image can be denoised by reconstructing image only with its sparse components.The over-complete dictionary with the vein features of finger vein images was constructed by using the Gaussian function. The performance of the new method was verified by both synthetic and real finger vein images. Experimental results show that this algorithm can get better PSNR 1-2 dB compared with the traditional denoising.