The image taken by the unmanned aerial vehicle (UAV) patrol power channel line has more salt and pepper noise. For the shortcoming of traditional median filtering and adaptive median filtering algorithm to filter out the high concentration of salt and pepper noise and preserve the image edge details, a median filtering algorithm based on partial differential is proposed. The algorithm analyzes whether the original pixel point is a suspected noise point by calculating whether the derivative value of the four pixels in the four directions from the original pixel point is greater than a set threshold, and if one of the derivative value is greater than the set threshold, it is determined to be a suspicious noise point, and the window is moved to if the derivative value is larger than the threshold, it is determined as a noise point. After the noise point is determined, the weighted convolution is performed with the field window of the mask and the noise point that has been set to obtain a new pixel value. The original noise point is replaced by a new pixel value. The final result shows that the algorithm has higher peak signal noise than that of the adaptive median filtering algorithm, and the denoising effect is better.