Abstract: To remove the impulse noise from surface defect images of strip steel, a threshold-based method of switching median filtering is presented according to the characteristics of impulse noise and surface defect images of strip steel. In the method, whether the central pixel point in the filter window is a noise point is determined by setting the threshold data, then the median filtering process is done for the noise point. Lots of simulation results showed that the method is simple and applicable, and it is evidently better than other conventional filtering methods for removing details kept up. The method has a wide applicability to low noise pollution and is available to engineering applications. It was also found that the inspection of images and relevant details are of importance to several technological processes.