Hi Gert, Thanks for the reply. I'm cc-ing the list since others might be interested in the answers as well.
> This is great news. Particular as the auto correct works in RAW > compared to the lensfun module and thus improves the demosaic output > of AMAZE (as stated by it's designer). Indeed, when the correction becomes sufficiently precise significant improvements can be noticed in the AMAZE output. I've also worked with the lensfun module and moved it to an earlier position in the pipeline. It's promising since the residual chromatic aberrations after lensfun correction are really small for my lenses. Unfortunately the corrections that are computed by lensfun contain "fuzziness" (they are kind of non-continuous due to big rounding errors), and applying those on non-demozaicaized images makes it plainly visible. I'll be working on fixing that directly in lensfun later on probably. So, currently the best option is to use auto-correction and disable TCA in the lensfun module. > I've noticed that the original auto correct ca module had fitting > issues with Longitudinal Chromatic Aberrations. That type of images > usually required the lensfun module with ca settings obtained from > other images. How does your new module perform? The general idea for the fitting hasn't changed, but you should notice improvements because the method for the evaluation of color shifts has changed. The fitting is now computed based on better estimations, with proper weighting. For example, a large pseudo-uniform blue sky does not have a bad influence on shift estimations. About the fitting itself, the applied red and blue color shift are resolved as 2-dimensional polynomials of degree 4 (hard-coded but can be modified in the source). That is, a sum of x^n*y^m components with n + m <= 4. The previous implementation used max(n,m) < 4, which is actually more components but gave worse results (because of the way it was computed and because smoothing the fitting with lower degrees is needed too). And it is applied to each pixel separately (not to large blocks like previously), with sub-pixel interpolation. All in all, this means that the module should be able to cope with any kind of reasonably smooth chromatic drifts (the search is limited to a max of 8 pixels instead of 4 previously, and this could be lifted if iterative passes are added). It makes no assumption whatsoever on the geometry of the lens. > And would you consider adding a manual slider option or do > you use SVD for the fitting? I like the zero-configuration approach. If you refer to the possibility to fine-adjust the amount of correction with a fixed geometry, like with the lensfun approach, maybe wait a little, it might become reasonable to use lensfun earlier in the pipeline. Best, Stéphane ------------------------------------------------------------------------------ Learn Graph Databases - Download FREE O'Reilly Book "Graph Databases" is the definitive new guide to graph databases and their applications. Written by three acclaimed leaders in the field, this first edition is now available. Download your free book today! http://p.sf.net/sfu/NeoTech _______________________________________________ darktable-devel mailing list darktable-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/darktable-devel
