Bibliography

(eng): For image segmentation in the bioimaging field, the Otsu thresholding algorithm is very often the algorithm of choice. It's simple and fast algorithm. The drwaback of this algorithm is that it does not account for the image contents, and, in the bioimaging context, it often sets the threshold too high. In result, the contours of the resulting binary objects do not fully cover the original objects. An alternative option is represented by algorithms based on iterative optimization, such as the active contours, deformable models, etc. These algorithms are iterative and possibly rather computationally expensive. An interesting trade-off between the two approaches has been described in Snell et al: Segmentation and shape classification of nuclei in DAPI images. This method uses cost function that relates to the quality of resulting boundary. In the poster, an implementation of the Snell algorithm in the form of an ImageJ plugin was presented.