Am 07.06.2013 12:57, schrieb johannes hanika: > my 2ct: > > - underexposure + digital exposure correction is terrible wrt noise > > - i think our highlight reconstruction (this is about hardware clipped > pixels in the input) has some problems currently, if i find some time i > want to fix this. > > j.
According to my understanding we have two separate issues. 1) pixels that are clipped in raw. If not all channels are clipped we can reconstruct the lightness information but not the pixel color. Reconstruction in LCh, as offered by the highlight reconstruction module, only works for isolated highlights where surrounding pixels have valid color information - does not work for the typical larger clipped areas in sky. One idea could be a color reconstruction with a bilateral filter in Lab. We populate an <x,y,I> room with a-b-values from unclipped pixels and reconstruct clipped ones by interpolating with their (x,y,I) values. 2) pixels that are driven above L=100 in darktable, e.g. in the basecurve module. There is no convenient method ATM to bring them back into reasonable values. Shadows&Highlights would be a good candidate. But the current implementation clamps all pixel values to [0,100] and loses highlight detail. This behavior could be changed. We only need to make sure that legacy history stacks deliver the current output. my 2 cents. Ulrich ------------------------------------------------------------------------------ How ServiceNow helps IT people transform IT departments: 1. A cloud service to automate IT design, transition and operations 2. Dashboards that offer high-level views of enterprise services 3. A single system of record for all IT processes http://p.sf.net/sfu/servicenow-d2d-j _______________________________________________ Darktable-users mailing list Darktable-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/darktable-users
