Abstract: A method for the feature extraction of flaws in magnetic flux leakage data of pipeline is proposed in this paper. The method takes the data from the inspection device of the pipeline as the data for feature extraction. The data is converted into gray image by the relation of magnetic density and pixel, and then the image is filtered. Then, the possible positions of pipeline flaws are found by simple judgment. Binary process of the image by using the suitable threshold is applied to grayscale images. The pipeline flaws' boundary is determined by the information of connectivity and chain code, and the type of flaws is also determined according to the sensors of inspection device.