Alberto Griggio
1467b858c5
fixed missing memory deallocation
2017-11-17 15:34:48 +01:00
Alberto Griggio
b25bac8c53
improved performance of Fattal in dcrop
...
Now only Fattal works on the full image, the rest of the pipeline (including denoising) always operates only on the visible crop
2017-11-17 15:27:56 +01:00
Alberto Griggio
339c794359
Fixed bug (typo) in LCP-based CA correction (now reenabled)
2017-11-16 21:22:17 +01:00
Alberto Griggio
619b3e9c63
fixed off-by-one error leading to segfault
2017-11-16 17:35:21 +01:00
heckflosse
652044e863
Don't create panasonic bithuff when panasonic_load_raw() is not used
2017-11-15 22:33:06 +01:00
heckflosse
3f1e47d3d5
Merge branch 'dev' of https://github.com/Beep6581/RawTherapee into dev
2017-11-15 14:57:28 +01:00
heckflosse
76ac9b0288
Fix for ph1_bithuff_t, thanks to Flössie
2017-11-15 14:56:57 +01:00
Beep6581
696299be2e
Merge pull request #4181 from Benitoite/patch-3
...
Enable detection of non-Apple clang Clang features
2017-11-15 08:08:07 +01:00
heckflosse
7f9608eea2
Merge branch 'dev' of https://github.com/Beep6581/RawTherapee into dev
2017-11-15 00:54:34 +01:00
heckflosse
18f775d805
review of ph1_bithuff
2017-11-15 00:54:05 +01:00
Alberto Griggio
19281eedd1
merge branch 'lensprofile-ca-correction' into 'dev'
2017-11-14 22:53:27 +01:00
Alberto Griggio
99f41baf4f
disable LCP-based CA correction, as it is currently broken
2017-11-14 22:52:57 +01:00
Flössie
02d6187cd8
procparams.h cleanup - Part 1
...
TODO:
- Part 2 (obviously 😉 )
- Remove `setDefaults()` in favour of c'tor
2017-11-14 21:20:07 +01:00
Flössie
aa414fca41
Clean rtengine::procparams::Threshold
2017-11-13 22:00:27 +01:00
heckflosse
0efee599a1
Fix bugs I introduced with the speedup for Phase One decoder
2017-11-13 18:57:44 +01:00
heckflosse
7507b74d6f
explicitly set -ftree-vectorize to get auto vectorizations even for builds using -o2 instead of -o3 (some distros do that for whatever reason)
2017-11-13 18:56:18 +01:00
Benitoite
46e4dece2e
Enable detection of non-Apple clang Clang features
...
Ups macOS cmake requirement to 3.3 and sets policy 25, to differentiate compiler features between AppleClang and plain ole' clang.
2017-11-13 09:51:21 -08:00
heckflosse
bd9592cb1f
Fix two warnings when using gcc < 7
2017-11-12 13:42:25 +01:00
Flössie
c702e5d847
Additional manual cleanups to ProcParams::(save|load)()
2017-11-11 21:11:57 +01:00
Flössie
157b42cd83
Introduce assignFromKeyfile() and saveToKeyfile()
2017-11-11 20:25:13 +01:00
Ingo Weyrich
efcdb561d2
Merge pull request #4177 from Beep6581/ph1speedup
...
Speedup for loading of Phase One files
2017-11-11 18:50:09 +01:00
heckflosse
4754d5a6cc
Speedup for loading of Phase One files, fixes #4050
2017-11-11 18:48:44 +01:00
heckflosse
3b36890546
Merge branch 'dev' into ph1speedup
2017-11-11 18:34:09 +01:00
heckflosse
8fa477793d
Speedup for initial thumb creation from raw files, fixes #4109
2017-11-11 18:31:37 +01:00
Ingo Weyrich
5946a00511
Merge pull request #4176 from Beep6581/extraclean
...
get rt cleaner when built using -Wextra #4155
2017-11-11 18:06:00 +01:00
Alberto Griggio
2cdc5fc69a
move Fattal before transform
2017-11-11 16:18:38 +01:00
heckflosse
67b6aec64e
Fattal: one less buffer alloc/dealloc
2017-11-09 13:06:12 +01:00
Alberto Griggio
b0ebcc30eb
Fattal: reduced number of temporary buffers needed
2017-11-08 21:49:50 +01:00
heckflosse
cb735125a2
Removed two stopwatches
2017-11-08 19:09:06 +01:00
heckflosse
2ff9ca0145
lot of small speedups for fattal
2017-11-08 19:01:52 +01:00
Alberto Griggio
75405404a9
Fattal: speed up FFT computation by rescaling the images to FFTW-friendly dimensions
2017-11-08 18:07:03 +01:00
Alberto Griggio
b8ff601efa
fixed silly typo in Median_Denoise, leading to segfault
2017-11-07 21:07:29 +01:00
Alberto Griggio
b1938e272c
added a dedicated Median_Denoise function that applies the median filter only on dark pixels
...
This gives a slight performance improvement for Fattal
2017-11-07 18:00:00 +01:00
Alberto Griggio
31520c1a74
remove "(Fattal02)" from the name of the HDR Tone Mapping tool
2017-11-07 11:18:07 +01:00
Alberto Griggio
4d010f8dec
do use a temporary image for Fattal in dcrop when it is needed
...
Because, sometimes it is needed indeed...
2017-11-07 10:31:03 +01:00
Alberto Griggio
6e17e38136
no need to use a temporary image for Fattal in dcrop
2017-11-06 23:00:52 +01:00
Alberto Griggio
d37f82498b
Fattal: fixed typo in assertion, and added one more assertion
2017-11-06 22:48:40 +01:00
heckflosse
11f756239b
Added comment to findMaxMinPercentile, interpolated minLum and maxLum. Removed some stop watches
2017-11-06 20:53:36 +01:00
heckflosse
81fbf0c2e1
Applied patch from @agriggio to check multithreading in fftw3
2017-11-06 19:09:50 +01:00
heckflosse
d810de4445
Double speed for rescale_bilinear
2017-11-06 19:08:30 +01:00
heckflosse
356d495252
Merge branch 'hdr_tonemap_fattal' of https://github.com/Beep6581/RawTherapee into hdr_tonemap_fattal
2017-11-06 01:12:37 +01:00
heckflosse
ab061283a6
fattal, about 100x speedup for findMaxMinPercentile()
2017-11-06 01:12:15 +01:00
Alberto Griggio
b640a37d79
Fattal GUI: set a lower bound of 0 for amount
2017-11-05 21:49:38 +01:00
Alberto Griggio
1a2b7d2cde
Fattal: extract luminance using the current working space matrix, not the sRGB one
...
Doesn't seem to make any practical difference though...
2017-11-05 21:32:06 +01:00
heckflosse
facb37be91
Speedup for exp-loop
2017-11-05 16:47:22 +01:00
heckflosse
1e0c5bfb5e
solved merge conflict
2017-11-05 16:29:59 +01:00
heckflosse
59043cc978
Speedup for log-loop in tmo_fattal02
2017-11-05 16:25:13 +01:00
Alberto Griggio
d187c2e20b
Fattal: raised threshold on black pixels clipping to avoid pixel artifacts
2017-11-05 16:05:50 +01:00
Alberto Griggio
d7136fc668
Fattal: apply a median filter on luminance on the deep shadows, to avoid boosting noise
...
Now preview matches output also for noisy images
2017-11-05 15:09:22 +01:00
Alberto Griggio
54783f6e2f
Fixed bug (uninitialized values) in ImProcFunctions::Median_Denoise
...
The code was not copying border pixels from medianOut to dst, leading to the
use of uninitialized values. This was not visible in RT because Median_Denoise
was always called with src == dst, so not copying the border causes no harm
2017-11-05 15:07:03 +01:00
