IFOC: Intensity Fitting on Overlapping Cover for Image Segmentation
Region-based image segmentation methods often require a global statistical description of the image intensities in the entire foreground and background, which however is in general not available for real world images due to the complicated intensity variation within the regions of interest. In this paper, we propose a foreground-background segmentation algorithm for images with intensity inhomogeneity in a level set framework, which exploits simple distribution of local image intensities. We assume that the intensities of the foreground and background within a small enough neighborhood are separable and can be well approximated by two constants. We call such a neighborhood an intensity separable neighborhood (ISN). Given a set of overlapping ISNs that form a cover of the entire image domain or a region of interest, we formulate image segmentation as a problem of seeking for an optimal level set function that represents the foreground and background with its positive and negative sign, and a pair of constants that approximate the local foreground and background intensities within the ISNs. This formulation is a significant extension and improvement of our previous work. The main contributions in this paper include: (1) We eliminate an intrinsic drawback in our previous work that the fitting functions are not well defined for points far away from the zero level set, which causes unstable performance of segmentation; (2) The new algorithm is more efficient than our previous algorithm due to the sparse placement of the ISNs.
KeywordsLevel set Overlapping neighborhoods Image segmentation Intensity inhomogeneity
- 2.Li, C., Kao, C., Gore, J., Ding, Z.: Implicit active contours driven by local binary fitting energy. In: Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (CVPR), pp. 1–7. IEEE Computer Society, Washington (2007)Google Scholar
- 4.Li, C., Xu, C., Gui, C., Fox, M.: Level set evolution without re-initialization: a new variational formulation. In: IEEE Computer Society Conference on Computer Vision and Pattern Recognition, CVPR 2005, vol. 1, pp. 430–436. IEEE (2005)Google Scholar