Scale space filtering is a well known approach in image segmentation. This approach determines precisely the thresholds in order to produce the image segmentation by histogram multi-thresholding. The main drawback of this approach is that it produces a coarse segmentation because it does not take into account the spatial arrangement of the pixels in the image and so, requires a second stage in order to obtain a finer segmentation. In this paper, we propose to associate a new criterion, the compactness degree, with the scale space filter in order to produce the segmentation by means of one single stage. For this purpose, this criterion exploits the connectedness and the homogeneity properties of pixels.
L. Busin, N. Vandenbroucke, L. Macaire, J.-G. Postaire, "Scale space filter based on homogeneity degree for color image segmentation" in Proc. IS&T CGIV 2006 3rd European Conf. on Colour in Graphics, Imaging, and Vision, 2006, pp 12 - 17, https://doi.org/10.2352/CGIV.2006.3.1.art00003