A Bi-level Image Segmentation Framework Using Gradient Ascent
In order to solve the problem of under-segmentation in traditional superpixel methods, a new image segmentation framework is proposed, which is based on gradient ascent including Simple Linear Iterative Clustering (SLIC) superpixels and watershed algorithm. First, SLIC method is adopted to generate uniform superpixels, which are then determined whether under-segmentation occurs by a homogeneity criterion. In heterogeneous regions, an adaptive watershed algorithm processes a more precise division based on luminance histogram. Experimental results show that the bi-level framework has good performance on detail-rich regions, without significantly increasing the time complexity compared with conventional SLIC.
KeywordsSegmentation Superpixel Watershed Subdivision
This work is supported by the National Natural Science Foundation of China (61571346). The research is also supported by the Fundamental Research Funds for the Central Universities and the Innovation Fund of Xidian University.
- 3.Boemer, F., Ratner, E., Lendasse, A.: Parameter-free image segmentation with SLIC. Neurocomputing 277(2018), 228–236 (2017)Google Scholar
- 6.Malik, J.: Learning a classification model for segmentation. In; ICCV, vol. 1, pp. 10–17 (2003)Google Scholar
- 7.Martin, D.R., Fowlkes, C., Tal, D., Malik, J.: A database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics. In: Proceedings of International Conference on Computer Vision, vol. 2, no. 11, pp. 416–423 (2001)Google Scholar