I agree there are many solutions to an image like this and that a watershed / level-set approach could work very well. I think it may work even better if you apply it to the ridge-enhanced filter response that Nessys uses rather than an isotropic Gaussian convolution filter.
If you look into the term “steerable filters”, you will see that the original filter design is based on the combination of convolving with a Gaussian followed by a derivative. It turns out that you combine both operations into a single convolution operation that can be interpolated in terms of filter angle. In the case of a ridge filter, even (2nd, 4th, etc.) derivatives are used.
This is the classic paper that coined the term:
The design and use of steerable filters
W.T. Freeman ; E.H. Adelson
The authors present an efficient architecture to synthesize filters of arbitrary orientations from linear combinations of basis filters, allowing one to adaptively steer a filter to any orientation, and to determine analytically the filter output as a function of orientation. Steerable filters may be designed in quadrature pairs to allow adaptive control over phase as well as orientation. The authors show how to design and steer the filters and present examples of their use in the analysis of orientation and phase, angularly adaptive filtering, edge detection, and shape from shading. One can also build a self-similar steerable pyramid representation. The same concepts can be generalized to the design of 3-D steerable filters.
The algorithm we have been specifically discussing here that is used in Nessys is based on the following paper:
Design of steerable filters for feature detection using canny-like criteria
M. Jacob ; M. Unser
We propose a general approach for the design of 2D feature detectors from a class of steerable functions based on the optimization of a Canny-like criterion. In contrast with previous computational designs, our approach is truly 2D and provides filters that have closed-form expressions. It also yields operators that have a better orientation selectivity than the classical gradient or Hessian-based detectors. We illustrate the method with the design of operators for edge and ridge detection. We present some experimental results that demonstrate the performance improvement of these new feature detectors. We propose computationally efficient local optimization algorithms for the estimation of feature orientation. We also introduce the notion of shape-adaptable feature detection and use it for the detection of image corners.
A more recent review and generalization of the concept can be found here:
A Unifying Parametric Framework for 2D Steerable Wavelet Transforms
Michael Unser and Nicolas Chenouard
Read More: https://epubs.siam.org/doi/abs/10.1137/120866014
Anyways, I think we are veering off topic. Please post a new topic when you have implemented your method and then we can discuss the comparison.