Unsupervised segmentation evaluation: an edge-based method
- 192 Downloads
Unsupervised segmentation evaluation method quantifies the quality of segmentation without the reference segmentation or user assistance. Although some methods have been proposed to statistically analyze the pixel values, these methods are not sensitive enough to provide a metric of segmentation quality. This paper uses the image edge, a more robust feature, to measure the quality of segmentation. An edge-based segmentation evaluation method is introduced in this paper, which can be applied to both image and single region segmentation evaluation. The proposed method evaluates the quality of segmentation with three edge-based measures: the edge fitness, the intra-region edge error, and the out-of-bound error. These measures encourage the outline of segmentation to align with the edge and punish the segmentation that exceeds the edge. Experiments results show that our method is more sensitive to under-segmentation and over-segmentation. Using the parameters optimized by the proposed method, the segmentation produced by the classic region growing method is visually similar to the state-of-the-art segmentation method.
KeywordsImage segmentation Objective evaluation Unsupervised evaluation
This work is supported by National Natural Science Foundation of China (61170193, 61370102, 61370185), Guangdong Natural Science Foundation (S2012010009865, S2012020011081, S2013010013432, S2013010015940, 2014A030306050), Science and Technology Planning Project of Guangdong Province (2011B090400041, 2012B010100039, 2012B040305011, 2012B010100040), Education and Science Programs of Guangdong Province (11JXZ012,14JXN065), Guangdong Higher Education Discipline and Profession Special Fund Projects(2013KJCX0174), Science and Technology Planning Project of Huizhou City (2011P002, 2011g012, 2011P005, 2011P003, 2011g011, 2013B020015008, 2014B020004026) and the Fundamental Research Funds for the Central Universities, SCUT (2015PT022).
- 5.Chen HC, Wang SJ (2004) The use of visible color difference in the quantitative evaluation of color image segmentation. In: IEEE international conference on acoustics, speech, and signal processing, 2004. Proceedings (ICASSP’04), vol 3. IEEE, pp iii–593Google Scholar
- 7.Donoser M, Schmalstieg D (2014) Discrete-continuous gradient orientation estimation for faster image segmentation. In: IEEE conference on computer vision and pattern recognition (CVPR), 2014, IEEE, pp 3158–3165Google Scholar
- 10.Maire M, Arbeláez P, Fowlkes C, Malik J (2008) Using contours to detect and localize junctions in natural images. In: IEEE conference on computer vision and pattern recognition, 2008. CVPR 2008. IEEE, pp 1–8Google Scholar
- 11.Malisiewicz T, Efros AA (2007) Improving spatial support for objects via multiple segmentations. British Machine Vision Conference (BMVC)Google Scholar
- 12.Meila M (2005) Comparing clusterings: an axiomatic view. In: Proceedings of the 22nd international conference on machine learning (ICML-05), pp 577–584Google Scholar
- 14.Miao Q, Cao Y, Xia G, Gong M, Liu J, Song J (2015) Rboost: label noise-robust boosting algorithm based on a nonconvex loss function and the numerically stable base learners. IEEE Trans Neural Netw Learn Syst PP(99):1–1Google Scholar
- 16.Sobel I, Feldman G (1968) A 3 × 3 isotropic gradient operator for image processing. In: a talk at the Stanford Artificial Project in, pp 271–272Google Scholar
- 18.Zhang H, Fritts JE, Goldman SA (2003) An entropy-based objective evaluation method for image segmentation. In: Electronic imaging 2004, international society for optics and photonics, pp 38–49Google Scholar