Multiscale line segment detector for robust and accurate SfM

Yohann Salaün (1, 2) , Renaud Marlet (2, 3, 4) , Pascal Monasse (4)

1 Ecole Centrale Paris
2 imagine [Marne-la-Vallée]
3 LIGM - Laboratoire d'Informatique Gaspard-Monge
4 ENPC - École des Ponts ParisTech

We propose a multiscale extension of a well-known line segment detector, LSD. We show that its multiscale nature makes it much less prone to over-segmentation, more robust to low contrast and less sensitive to noise, while keeping the parameterless advantage of LSD and still being fast. Moreover, we show that in scenes with little or no feature points, but where it is however possible to perform structure from motion from matched line segments, the accuracy is significantly improved. This provides an objective and automatic quantitative assessment of our detector that goes much beyond the usual qualitative visual inspection found in the literature.

Domains: Computer Vision and Pattern Recognition [cs.CV]

Yohann Salaün, Renaud Marlet, Pascal Monasse. Multiscale line segment detector for robust and accurate SfM. International Conference on Pattern Recognition, Dec 2016, Cancun, Mexico. pp.2001-2006. ⟨hal-01416342⟩