So, today the watersheds approach can be used as initially intended. It was discussed and explained many times […references to be included..]. Image gradient is used to find minima as markers of the “particles” of interest, and the segmentation is build as watershed line between these markers. I will put here just a very simple sketch to illustrate the idea one more time. Fig. 1 demonstrates the series of images in 2D in the first row, the profiles of these images in the second row, and the result in 3D in the third row.
Fig.1. The idea of image segmentation using watersheds.
The purpose of the images included in Fig.2 is to show the adaptability of the algorithm. Changes in color of layers show quite well rather good adaptation of the algorithm to variations of the intensity of staining from one section to another. Despite different colors, varying from red to orange to yellow, the structure of layers remains preserved.
Very interesting article. Thanks.