Conclusion#
We have seen that segmentation consists in dividing an image into several homogeneous regions. The homogeneity of a region is based on color, texture, contours, etc.
The methods of segmentation are very diverse, and we have only seen a few of them. Other existing methods include active contours or the so-called snakes, level sets, Markovian models, etc. In addition to these “model-based” methods, deep learning allows significant increases in the results.
At last, we list some criteria to evaluate the quality of a segmentation method. These criteria are useful to compare different segmentation methods, as presented in Lab 4.
References#
J.B. MacQueen, “Some Methods for classification and Analysis of Multivariate Observations”, Proceedings of 5th Berkeley Symposium on Mathematical Statistics and Probability, vol. 1, p. 281–297, 1967
N. Otsu, “A threshold selection method from gray-level histograms”, IEEE Transactions on Systems, Man, and Cybernetics, vol. 9, no 1, p. 62-66, 1979.
H. Steinhaus, “Sur la division des corps matériels en parties”, Bull. Acad. Polon. Sci., vol. 4, no 12, p. 801-804, 1957.