The extraction of the line features is very critical for traffic sign recognition. Aiming at the drawbacks of existing Hough transform,such as peak diffusion in a parameter space,heavy computational complexity and massive storage requirements,an adaptive many-to-many mapping scheme insensitive to the threshold is proposed to exactly extrac the lines from a image with low computational complexity. Experimental results reveal that the range of selectable threshold of the new method is three times that of MTO_θ,while the running time is 10% that of MTO_ρ. These results reported here show that proposed method is more efficient for extracting the lines in traffic sign images with acceptable running time,and it can also be used for line detection in symbol images and bill images.