DC-Seg: Hierarchical Image Segmentation


In this work, we propose an image segmentation method, that returns a hierarchy of segments with differing granularity. We propose an efficient discrete-continuous optimization of oriented gradient signals, that are passed to an oriented watershed transformation to build a hierarchical segment tree as output. We are able to provide segmentation performance competitive to state-of-the-art (even without any additional spectral analysis) while reducing computation time by a factor of 40. Additionally, since we do not have to apply a spectral analysis, we are able to reduce the memory profile by a factor of 10. In such a way, our segmentation method allows to provide useful input to various computer vision applications in short computation time.

Input Image 1

Figure 1: Comparison of image segmentation results on exemplary image from BSDS 500.
One level of the obtained hierarchy is shown.Segments are mapped to their mean RGB value.
Left: input image, Middle: Ultra Contour Map (UCM) [1] result, Right: DC-Seg Result
UCM required 179 seconds and 2.4 gigabyte RAM, whereas DC-Seg provides results in 5 seconds using only 128 megabytes RAM.

Input Image 1

Figure 2: Comparison to Ultra Contour Map (UCM) [1] on BSDS 500.
[1] Contour detection and hierarchical image segmentation. Arbelaez et al. T-PAMI 2010.


We provide an implementation of our method in Matlab. The code contains all scripts for evaluation on BSDS 500 and examples for segmenting individual color images: Download CODE Ver.1 (Matlab).

When using this software, please acknowledge the effort that went into development by referencing the paper!


  1. Discrete-Continuous Gradient Orientation Estimation for Faster Image Segmentation (PDF)
    Michael Donoser and Dieter Schmalstieg
    Proceedings of Conference on Computer Vision and Pattern Recognition (CVPR), 2014