Hola y bienvenido a PixInsight Forum.
Dada la importancia del tema al que se refiere tu mensaje, voy a continuar en Inglés para que todos los usuarios puedan participar.
I would not trust these results, especially the filamentary structures shown around the nucleus. They are "correct" in the mathematical sense, but there is no guarantee that they represent actual physical structures. I hope you won't mind if I include one of your images that best show this potential problem:
The multiscale median transform algorithm (MMT) uses morphological median filters to build a multiscale representation of the image. A discrete median filter uses a structuring element to define a neighborhood around each pixel. The shape and orientation of this structuring element is very important because it defines the fileter's morphological behavior. For example, you can see how median filters and structuring elements work in practice if you play with the MorphologicalTransformation tool in PixInsight.
In the original MMT algorithm,
[1][2] square structuring elements are used. This shape is very problematic for image processing applications of MMT because it introduces a strong anisotropic behavior (for example, a rectangular structuring element tends to distort circular or elliptical shapes and smooth curves). In our implementation we have modified the original algorithm with multiway structures that approximate circular structuring elements. These elements are much better than the original square ones to represent smooth curves and round structures, such as stars and most deep-sky objects and image features (although they tend to generate small artifacts at sharp corners). As a result, our implementation has much better isotropy properties and preserves the original ringing-free behavior of MMT.
However, since we are working with digital filters and structuring elements have to be represented as discrete kernels, achieving a perfectly isotropic median filter is impossible. When you apply MMT to perform aggressive transformations, the lack of isotropy can become problematic. This problem is even worse with HDRMultiscaleTransform because it applies MMT iteratively. For these comet images, where you have to apply an extremely strong dynamic range compression, I would stick with linear, purely isotropic transforms (that is, the starlet transform, aka
à trous transform, instead of MMT).
_______________________
[1] Starck, J.-L., Murtagh, F. and J. Fadili, A. (2010),
Sparse Image and Signal Processing: Wavelets, Curvelets, Morphological Diversity, Cambridge University Press.
[2] Barth, Timothy J., Chan, Tony, Haimes, Robert (Eds.) (2002),
Multiscale and Multiresolution Methods: Theory and Applications, Springer. invited paper: Jean-Luc Starck,
Nonlinear Multiscale Transforms, pp. 239-279.
