Human vision inspired multi-scale line segments merging and filtering

A novel method for merging and filtering line segments which is designed based on line segments detector (LSD) is proposed. Compared to state-of-the-art, our algorithm reduces the complexity of the algorithm by implementing the property of cross product between two vectors, thus improves the speed. And inspired by human vision, a further multi-scale filtering step is conducted to capture predominant line segments and leave out trivial segments in different scaled images. Experiments on both synthetic and real images show that the merging and filtering line segments are closer to human-marked ground-truth and consistent with that of human vision system.