Efficient Hierarchical Graph-based Video Segmentation ------------------------------------------------ Last updated date: 2014/9/7 For research purposes, using or modifying the soure code is granted, but any form of commercial usage is not allowed. ------------------------------------------------ Contact info: Please e-mail author, Haw-Shiuan Chang (samkendi@hotmail.com), if you find any bug in our code or have any question to this code. ------------------------------------------------ About: This code is the accelerated version of the GBH [1] implemented from LIBSVX v3.0 (http://www.cse.buffalo.edu/~jcorso/r/supervoxels/) [2]. Based on my testing, this code consuming only 1/10 computational time and memory of the implementation from LIBSVX on normal paremeter settings. We accelerate the code in order to extend our hierarchical image segmentation work [3] in the future. If you use this code and publish your results, please cite [1], [2] and [3]. ------------------------------------------------ Notice: 1. The I/O part of the code only supports Windows platform for now. It should be fairly easy to modify the I/O function back to linux platform as that in LIBSVX. 2. The segmentation results might be a little bit different from that generated by LIBSVX v3.0 because the contour strength between segments might be the same and we might have different merging orders for those segments. ------------------------------------------------ Usage: Please refer to chapter 2.2.1 in the LIBSVX tutorial file (tutorial.pdf). ------------------------------------------------ [1] Matthias Grundmann, Vivek Kwatra, Mei Han and Irfan Essa1, "Efficient hierarchical graph-based video segmentation," CVPR, 2010. [2] Chenliang Xu and Jason Corso, "Evaluation of super-voxel methods for early video processing," CVPR, 2012. [3] Haw-Shiuan Chang and Yu-Chiang Frank Wang, "Simple-to-Complex Discriminative Clustering for Hierarchical Image Segmentation," ACCV, 2014.