eduzhai > Applied Sciences > Engineering >

Spectral Image Segmentation with Global Appearance Modeling

  • Save

... pages left unread,continue reading

Document pages: 10 pages

Abstract: We introduce a new spectral method for image segmentation that incorporateslong range relationships for global appearance modeling. The approach combinestwo different graphs, one is a sparse graph that captures spatial relationshipsbetween nearby pixels and another is a dense graph that captures pairwisesimilarity between all pairs of pixels. We extend the spectral method forNormalized Cuts to this setting by combining the transition matrices of Markovchains associated with each graph. We also derive an efficient method that usesimportance sampling for sparsifying the dense graph of appearancerelationships. This leads to a practical algorithm for segmentinghigh-resolution images. The resulting method can segment challenging imageswithout any filtering or pre-processing.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×