…ftware

Works better than legacy gamma for high dynamic range (> 10 EV)
BREAKS THE PREVIOUS VERSION AND I DON'T KNOW WHY

…le OpenCL

…ctor from optimization

…els are < 0. and exit the optimization

…nning

…lver

…ard without solving equations on paper first -_-\nfrom the maths, remove redondant GUI sliders of dependant variables

…mizer. change the defaults settings so avoid tone-mapping by default

add a color picker for the black input
remove the noise/black correction to avoid inaccuracies at low dynamic range - the black level has to be absolutely set in exposure module. Now the log functions will only be thresholded.

…non-standard log coeff

… artefacts in low-right.

…ns of the log

…ute the lightness

Clamp input values after dividing by grey, which may reduce the noise amplification in log2
Correct the black color picker to take the min (not the max)
Still need to scream when negative pixels are present

Make the optimizers more robust and consistent
Correct the UI labels

Use average color in black color-picker to reduce the noise detection

Add thread_mutex_lock around GTK events

This reverts commit 539ae88.

…nto color-grading

Speed-up the CPU by 1.8× improving vectorization and using an approximate fast log2 version
Note : OpenCl version stays with exact vectorized log2 as it makes no speed difference

…ftware

Works better than legacy gamma for high dynamic range (> 10 EV)
BREAKS THE PREVIOUS VERSION AND I DON'T KNOW WHY

Get rid of the previous code, make things more simple and try to enable OpenCL

Change to log profile, add color sampling and best  parameters guessing

improve the auto tuner, refine the noise handling

uncomment the optimization loop

improve the solver, add a black correction slider

improve the GUI, change the default settings, remove the linearity factor from optimization

comment the loop now useless

clamp output and fix GUI glitches.\n Add a error log message when pixels are < 0. and exit the optimization

fix the GUI glitches at module activation

change the black/noise level auto estimation

prevent the update of the parameters if there is already an update running

reinclude the old gamma version and add legacx params support

add the log profile function, but DOES NOT WORK YET !

clang-format

improve the auto-optimizer, do the computations in RGB, remove the solver

correct important maths mistakes\nI shouldn't be allowed near a keyboard without solving equations on paper first -_-\nfrom the maths, remove redondant GUI sliders of dependant variables

(hopefully) fix once for all the slider glitches while using the optimizer. change the defaults settings so avoid tone-mapping by default

add a color picker for the black input
remove the noise/black correction to avoid inaccuracies at low dynamic range - the black level has to be absolutely set in exposure module. Now the log functions will only be thresholded.

…non-standard log coeff

… artefacts in low-right.

…ns of the log

…ute the lightness

Clamp input values after dividing by grey, which may reduce the noise amplification in log2
Correct the black color picker to take the min (not the max)
Still need to scream when negative pixels are present

Make the optimizers more robust and consistent
Correct the UI labels

Use average color in black color-picker to reduce the noise detection

Add thread_mutex_lock around GTK events

Revert "Add an SSE2 version which is not significantly faster"

This reverts commit 539ae88.

Speed-up the CPU by 1.8× improving vectorization and using an approximate fast log2 version
Note : OpenCl version stays with exact vectorized log2 as it makes no speed difference

…ktable into color-grading