* Enable dynamic range - settings normalize luminance
* Change range DR sigma offset
* improve normalize_mean_dt with mean std before
* Save work on normalize dynamic range
* Improve normalize for TM
* Added LA retinex to normalize_mean
* Clean and optimize code
* Disabled normalize DR - improve TM and Retinex
* Clean code - first part
* Clean code - part two
* Change 2 default values Retinex
* Change order TM and Retinex in process
* remove the order number in - add tool to current spot -
* Fixed bug in TM
* Some change Retinex - default...
* optimize mean sigma
* cleanup for mean_sig
* Fix some cppcheck issues
* convert one large loop into three loops to avoid performance penalty on machines with 4-way L1 cache
Co-authored-by: Ingo Weyrich <heckflosse67@gmx.de>
* Remove unfilled formatting placeholders in KLT parser.
* Fix presumably unintentionally missing reference on PlanatPtr and ChunkyPtr call operators.
* Fix catching pointer and make use-after-free of workimg easier to spot.
Make sure all of our thrown exceptions derive from std::exception and then catch
by reference instead of by pointer.
* Fix mismatch between array form new and non-array form delete.
* Simplify memory management of embedded color profiles by unifying allocation to use operator new.
Deleting a spot while hovering over one of the circles for adjusting the
radius or feathering could trigger a crash. This is because immediately
after deleting a geometry object, EditSubscriber::getCursor is given an
object ID based on old data, causing the spot tool to attempt to get
information about the deleted geometry object.
This commit fixes the issue by returning the default cursor if the
active spot is -1. It works because the active spot is set to -1 when
deleting a spot, and there is no need to use a special cursor when there
is no active spot.
A better long-term solution would be updating the geometry data before
calling EditSubscriber::getCursor.
* Added Guidedfilter to avoid color shift
* Various improvment to avoid color shift
* Small change to mint maxt
* local adjustments avoid color shift: reduce memory usage by width * height * 8 byte if Soft Radius > 0
* Remove StopWatch
* local adjustments avoid color shift: speedup for last loop
* cleanups
* one more cleanup
* Added checkbox Munsell correction only - uniform perceptual lab
* Refine some settings
* Clean-up - other small refinement
Co-authored-by: Ingo Weyrich <heckflosse67@gmx.de>
* [cmake] Switch to find_package for TIFF
- Included in cmake 3.5 and later
- Provides better configuration messages
* Add missing TIFF library to rtgui and rtexif
* Removed unnecessary quotes so that cmake sees the list as a list
* Add OpenMP to rtengine if OpenMP is enabled
