1Nankai University 2University of Oxford 3Tsinghua University 4UCSD
In this paper, we propose a real-time system, Hierarchical Feature Selection (HFS), that performs image segmentation at a speed of 50 frames-per-second. We make an attempt to improve the performance of previous image segmentation systems by focusing on two aspects: (1) a careful system implementation on modern GPUs for efficient feature computation; and (2) an effective hierarchical feature selection and fusion strategy with learning. Compared with classic segmentation algorithms, our system demonstrates its particular advantage in speed, with comparable results in segmentation quality. Adopting HFS in applications like salient object detection and object proposal generation results in a significant performance boost. Our proposed HFS system (will be open-sourced) can be used in a variety computer vision tasks that are built on top of image segmentation and superpixel extraction.
- HFS: Hierarchical Feature Selection for Efficient Image Segmentation. Ming-Ming Cheng, Yun Liu, Qibin Hou, Jiawang Bian, Philip Torr, Shi-Min Hu, Zhuowen Tu, ECCV 2016. [Project page][pdf][bib][C++]
Hierarchical Feature Selection (HFS)
Fig. 1: Pipeline of our methods.
Fig. 2: HFS for region merging
Evaluation on BSDS500 dataset
Fig. 3: Experimental evaluation for boundaries on BSDS500.
Fig. 4: Experimental evaluation for regions on BSDS500.
Fig. 5: Some examples of EGB, SLIC and our method. Left: source image. Middle left: SLIC . Middle right: EGB . Right: ours.
Links to most related works
 Slic superpixels compared to state-of-the-art superpixel methods. Achanta et al. IEEE TPAMI, 2012
 Contour detection and hierarchical image segmentation. Arbelaez et al. IEEE TPAMI, 2011
 Multiscale combinatorial grouping. Arbelaez et al. IEEE CVPR, 2014
 Mean shift: A robust approach toward feature space analysis. Comaniciu et al. IEEE TPAMI, 2002
 Spectral segmentation with multiscale graph decomposition. Cour et al. IEEE CVPR, 2005
 Ecient graph-based image segmentation. Felzenszwalb et al. IJCV, 2004
 Fast and eective l0 gradient minimization by region fusion. Nguyen et al. IEEE ICCV, 2015
 gslicr: Slic superpixels at over 250hz. Ren et al. arXiv, 2015
 Fast partitioning of vector-valued images. Storath et al. SIAM Journal on Imaging Sciences, 2014