2008 IEEE Conference on Computer Vision and Pattern Recognition
DOI: 10.1109/cvpr.2008.4587660
Full text: Download
We present a method for decomposing an image into its intrinsic reflectance and shading components. Differ- ent from previous work, our method examines texture in- formation to obtain constraints on reflectance among pixels that may be distant from one another in the image. We ob- serve that distinct points with the same intensity-normalized texture configuration generally have the same reflectance value. The separation of shading and reflectance compo- nents should thus be performed in a manner that guaran- tees these non-local constraints. We formulate intrinsic im- age decomposition by adding these non-local texture con- straints to the local derivative analysis employed in con- ventional techniques. Our results show a significant im- provement in performance, with better recovery of global reflectance and shading structure than by previous methods.