According to the characteristics of fingerprint images,the soft and hard threshold functions used in traditional wavelet threshold denoising method were improved and a polynomical fitting model was built to implement the adaptive selection of parameter α.Based on a large number of experimental data,and in the light of the fitting model α can be adaptively selected under different noise intensity and different wavelet decomposition levels.The computer simulation results show that the relative error of processed image PSNR between the results by using the model selected α and by using the optimal α is within 1%,and there appears a better visual effect.Furthermore,the method has a higher denoising efficiency and real-time processing ability compared to the soft and hard threshold methods with the increase of noise.