Color eq 2 #283
Color eq 2 #283
Conversation
libify the gamut LUT and gamut mapping methods
One that works, now.
|
|
Thank you ! We need to remove all |
|
|
||
| // fetch the corresponding y values | ||
| const float y_prev = gamut_lut[xi]; | ||
| const float y_next = gamut_lut[xii]; |
There was a problem hiding this comment.
Hi @aurelienpierre , i stumbled across this while looking for colorbalancergb issues both with opencl and cpu code. As we later do a simple lin approx i think we can/should avoid the division by (x_next - x_prev) and simply do
// fetch the corresponding y values
const float y_prev = gamut_lut[xi];
// return y_prev if we are on the same integer LUT element or do linear interpolation
return y_prev + ((xi != xii) ? (x_test - x_prev) * (gamut_lut[xii] - y_prev) : 0.0f);
There was a problem hiding this comment.
it's not a linear interpolation if you don't do
There was a problem hiding this comment.
Isn't dx here either 1 or zero due to floor/ceil above?
|
Removed my comments... |
Guided filter still produces NaN (then black squares) in high ISO pictures. Can't figure out why, mitigation is to raise the black point.