Title :
Functional vanishing point estimation via a filtered-Radon operator
Author :
Mantzel, William ; Romberg, Justin
Author_Institution :
Georgia Inst. of Technol., Atlanta, GA, USA
Abstract :
When available, vanishing points in a scene are a key factor in effectively recovering absolute camera orientation, thus simplifying the structure-from-motion problem. We present a novel method for estimating vanishing points without explicitly detecting line features. This approach first maps images into line-space with a filtered- Radon operator, allowing subtle line textures to contribute, and improving the angular resolution of broken or occluded segments of the same line. Then, we use a robust coarse-to-fine method to jointly estimate the three vanishing points. We evaluate our method on video sequences, demonstrating robustness to clutter lines as well as the ability to effectively utilize subtle edge-texture information.
Keywords :
Radon transforms; hidden feature removal; image sequences; image texture; video signal processing; camera orientation; edge-texture information; filtered-Radon operator; functional vanishing point estimation; occluded segments; subtle line textures; video sequences; Estimation; Feature extraction; Image edge detection; Joints; Optimization; Robustness; Transforms; Radon transform; absolute orientation; vanishing points;
Conference_Titel :
Image Processing (ICIP), 2010 17th IEEE International Conference on
Conference_Location :
Hong Kong
Print_ISBN :
978-1-4244-7992-4
Electronic_ISBN :
1522-4880
DOI :
10.1109/ICIP.2010.5651773