Abstract: In order to solve the problems of poor salt and pepper noise processing effect, poor real-time performance, edge loss and poor connection of strong and weak edges in weld image processing, an improved edge enhancement algorithm based on field programmable gate array (FPGA) is proposed. Firstly, the acquired video stream image is converted to gray level, and the adaptive median filter is used for noise reduction. Secondly, the image gradient vector is extracted by the 8-direction Sobel operator, and the adaptive threshold is used to improve the edge detection accuracy. Finally, the edge features are enhanced by morphological operation. The experimental results show that under the 100 MHz clock, it only takes 23.2016 ms to process a 768×1 024 weld image using FPGA, and the algorithm time is reduced to the millisecond level. The improved algorithm can effectively remove salt and pepper noise, filter false edges and retain real edge information.