Aiming at the outside impurities interference problems arising in the course for a round wire defect image denoising and feature extraction studies,based on the data collected from the field, the parameters and other information were used to model six surface defects common to the round wire.Three kinds of common image noise model were analyzed and compared by using a variety of filtering algorithms. By evaluating corresponding PSNR,appropriate pretreatment scheme was confirmed. The results show that the mean filtering algorithm has good effect on the Gaussian noise treatment, and the median filtering algorithm has good effect on the pretreatment of the salt-and-pepper noise, the mean filtering algorithm and median filtering algorithm have better processing results for Poisson’s noise; the SURF algorithm is used to extract the image feature. The conclusion provides reference for solving the problem of wire surface defects.