| Paper Details: | Downloads: 687 |
| Serial Number: | P1151151088
|
| Title: | Multi-view Video Segmentation based on Bayesian Estimation and Graph Cut
|
| Authors: | Anh Tu Tran, Koichi Harada
|
| Abstract: | In this paper, we propose a method, which requires no interactive operation, to segment human object from multi-view video. Our method consist of two stages: for initial frame of the video sequence, we automatically extract object based on saliency model and iterated Graph cut. After having segmented object in first frame, we propose the algorithm combining Bayesian estimation and minimizing energy function using graph cut to segment object. In our energy function, the color, depth and spatial-temporal coherence are integrated in data term. Smooth term is encoded the penalty cost of the neighboring pixels with different labels. By combining Bayesian estimation, minimizing energy functions via graph cut is speeded-up. Experiment results on test sequences are encouraging.
|
| Keywords: | Multi-views/Stereo Object segmentation, Automatic Object Segmentation, Object/Foreground extraction
|
| Journal/Conference: | International Journal of Graphics, Vision and Image Processing
|
| Volume: | 12
|
| Issue: | 1
|
| Submission Date: | 12/20/2011 12:00:00 AM
|
| Review Date: | 2/1/2012 12:00:00 AM
|
| Publishing Date: | 2/6/2012 12:00:00 AM
|
| Article Downloads: | 687
|
| Download: |
|