Extracting Lines in Noisy Image Using Directional Information

Detection of lines in noisy image is not easy. When using Hough transform, multiple false peaks may be generated from collinear noisy edge points, which in turn may create false line segments. To overcome this problem, we introduce a method for detecting lines in noise based on directional information. Orientations are generated by Gabor filters to guide an anisotropic Gaussian filtering process, and they are also used in the peak selection in Hough transform. The experimental results show the effectiveness of this method.