Radon sinogram decomposition for line segmentation

In this paper, a novel line segmentation algorithm is proposed as an extension to our previous works, i.e., line detection using multilayer fractional Fourier methods. First, a multilayer Fourier-based Radon transform (RT-MLFRFT) is applied to detect the location and the orientation of the straight lines. Afterwards, a set of windows-of-interest in spatial domain is designed to decompose the complex Radon space to a number of Radon spaces with a single transparent butterfly. This is to eliminate the effect of the neighbour RT peaks and guarantee the correct segment detection. Experimental results show that the proposed method enjoys superior performance compared with existing similar representative works.