Line in traffic sign image is a significant feature for traffic sign recognition. A modified Hough transform is proposed for line detection in traffic sign image. There exist some drawbacks in the existing Hough transform

such as peak diffusion in parameter space

heavy burden of computational complexity

and massive storage requirement. In this paper an adaptive many-to-many mapping scheme is proposed to overcome these drawbacks. With the new mapping scheme

the peak corresponding to an image line does not spread around in parameter space and is insensitive to parameter quantization

so the threshold can be selected conveniently. These characteristics are very helpful to accurately detect the parameter of a line in image. Experimental results reveal that the proposed method is more efficient for extracting the line in traffic sign image at the cost of an acceptable running time.