Changes to black compression and saturation controls. Black compression from 0-50 acts the same as 0-100 on the previous version, compressing dark tones without crushing blacks. 50-100 then starts crushing blacks until by 100 on the slider, all tones up to the set black point are sent to zero. In the new saturation control, negative values of the slider set a linear curve rather than an inverted S curve, and smoothly decrease saturation to zero across the board.

This commit is contained in:
Emil Martinec
2010-10-26 22:59:18 -05:00
commit 926056c2c2
620 changed files with 130476 additions and 0 deletions

221
rtengine/ipsharpen.cc Normal file
View File

@@ -0,0 +1,221 @@
/*
* This file is part of RawTherapee.
*
* Copyright (c) 2004-2010 Gabor Horvath <hgabor@rawtherapee.com>
*
* RawTherapee is free software: you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* RawTherapee is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with RawTherapee. If not, see <http://www.gnu.org/licenses/>.
*/
#include <rtengine.h>
#include <improcfun.h>
#ifdef _OPENMP
#include <omp.h>
#endif
#include <minmax.h>
#include <gauss.h>
#include <bilateral2.h>
namespace rtengine {
#undef CLIP
#undef CMAXVAL
#undef ABS
#define CMAXVAL 0xffff
#define CLIP(a) ((a)>0?((a)<CMAXVAL?(a):CMAXVAL):0)
#define ABS(a) ((a)<0?-(a):(a))
void ImProcFunctions::dcdamping (float** aI, unsigned short** aO, float damping, int W, int H) {
#ifdef _OPENMP
#pragma omp for
#endif
for (int i=0; i<H; i++)
for (int j=0; j<W; j++) {
float I = aI[i][j];
float O = (float)aO[i][j];
if (O==0.0 || I==0.0) {
aI[i][j] = 0.0;
continue;
}
float U = -(O * log(I/O) - I + O) * 2.0 / (damping*damping);
U = MIN(U,1.0);
U = U*U*U*U*(5.0-U*4.0);
aI[i][j] = (O - I) / I * U + 1.0;
}
}
void ImProcFunctions::deconvsharpening (LabImage* lab, unsigned short** b2) {
if (params->sharpening.enabled==false || params->sharpening.deconvamount<1)
return;
int W = lab->W, H = lab->H;
float** tmpI = new float*[H];
for (int i=0; i<H; i++) {
tmpI[i] = new float[W];
for (int j=0; j<W; j++)
tmpI[i][j] = (float)lab->L[i][j];
}
float** tmp = (float**)b2;
#ifdef _OPENMP
#pragma omp parallel
#endif
{
AlignedBuffer<double>* buffer = new AlignedBuffer<double> (MAX(W,H));
float damping = params->sharpening.deconvdamping / 5.0;
bool needdamp = params->sharpening.deconvdamping > 0;
for (int k=0; k<params->sharpening.deconviter; k++) {
// apply blur function (gaussian blur)
gaussHorizontal<float> (tmpI, tmp, buffer, W, H, params->sharpening.deconvradius / scale, multiThread);
gaussVertical<float> (tmp, tmp, buffer, W, H, params->sharpening.deconvradius / scale, multiThread);
if (!needdamp) {
#ifdef _OPENMP
#pragma omp for
#endif
for (int i=0; i<H; i++)
for (int j=0; j<W; j++)
if (tmp[i][j]>0)
tmp[i][j] = (float)lab->L[i][j] / tmp[i][j];
}
else
dcdamping (tmp, lab->L, damping, W, H);
gaussHorizontal<float> (tmp, tmp, buffer, W, H, params->sharpening.deconvradius / scale, multiThread);
gaussVertical<float> (tmp, tmp, buffer, W, H, params->sharpening.deconvradius / scale, multiThread);
#ifdef _OPENMP
#pragma omp for
#endif
for (int i=0; i<H; i++)
for (int j=0; j<W; j++)
tmpI[i][j] = tmpI[i][j] * tmp[i][j];
} // end for
delete buffer;
} // end parallel
#ifdef _OPENMP
#pragma omp for
#endif
for (int i=0; i<H; i++)
for (int j=0; j<W; j++)
lab->L[i][j] = lab->L[i][j]*(100-params->sharpening.deconvamount) / 100 + (int)CLIP(tmpI[i][j])*params->sharpening.deconvamount / 100;
for (int i=0; i<H; i++)
delete [] tmpI[i];
delete [] tmpI;
}
void ImProcFunctions::sharpening (LabImage* lab, unsigned short** b2) {
if (params->sharpening.method=="rld") {
deconvsharpening (lab, b2);
return;
}
if (params->sharpening.enabled==false || params->sharpening.amount<1 || lab->W<8 || lab->H<8)
return;
int W = lab->W, H = lab->H;
#ifdef _OPENMP
#pragma omp parallel
#endif
{
unsigned short** b3;
AlignedBuffer<double>* buffer = new AlignedBuffer<double> (MAX(W,H));
if (params->sharpening.edgesonly==false) {
gaussHorizontal<unsigned short> (lab->L, b2, buffer, W, H, params->sharpening.radius / scale, multiThread);
gaussVertical<unsigned short> (b2, b2, buffer, W, H, params->sharpening.radius / scale, multiThread);
}
else {
b3 = new unsigned short*[H];
for (int i=0; i<H; i++)
b3[i] = new unsigned short[W];
bilateral<unsigned short, unsigned int> (lab->L, (unsigned short**)b3, b2, W, H, params->sharpening.edges_radius / scale, params->sharpening.edges_tolerance, multiThread);
gaussHorizontal<unsigned short> (b3, b2, buffer, W, H, params->sharpening.radius / scale, multiThread);
gaussVertical<unsigned short> (b2, b2, buffer, W, H, params->sharpening.radius / scale, multiThread);
}
delete buffer;
unsigned short** base = lab->L;
if (params->sharpening.edgesonly)
base = b3;
if (params->sharpening.halocontrol==false) {
#pragma omp for
for (int i=0; i<H; i++)
for (int j=0; j<W; j++) {
int diff = base[i][j] - b2[i][j];
if (ABS(diff)>params->sharpening.threshold) {
int val = lab->L[i][j] + params->sharpening.amount * diff / 100;
lab->L[i][j] = CLIP(val);
}
}
}
else
sharpenHaloCtrl (lab, b2, base, W, H);
if (params->sharpening.edgesonly) {
for (int i=0; i<H; i++)
delete [] b3[i];
delete [] b3;
}
}
}
void ImProcFunctions::sharpenHaloCtrl (LabImage* lab, unsigned short** blurmap, unsigned short** base, int W, int H) {
int scale = 100 * (100-params->sharpening.halocontrol_amount);
unsigned short** nL = base;
#pragma omp parallel for if (multiThread)
for (int i=2; i<H-2; i++) {
int max1 = 0, max2 = 0, min1 = 0, min2 = 0, maxn, minn, np1, np2, np3, min, max;
for (int j=2; j<W-2; j++) {
int diff = base[i][j] - blurmap[i][j];
if (ABS(diff) > params->sharpening.threshold) {
// compute maximum/minimum in a delta environment
np1 = 2*(nL[i-2][j] + nL[i-2][j+1] + nL[i-2][j+2] + nL[i-1][j] + nL[i-1][j+1] + nL[i-1][j+2] + nL[i][j] + nL[i][j+1] + nL[i][j+2]) / 27 + nL[i-1][j+1] / 3;
np2 = 2*(nL[i-1][j] + nL[i-1][j+1] + nL[i-1][j+2] + nL[i][j] + nL[i][j+1] + nL[i][j+2] + nL[i+1][j] + nL[i+1][j+1] + nL[i+1][j+2]) / 27 + nL[i][j+1] / 3;
np3 = 2*(nL[i][j] + nL[i][j+1] + nL[i][j+2] + nL[i+1][j] + nL[i+1][j+1] + nL[i+1][j+2] + nL[i+2][j] + nL[i+2][j+1] + nL[i+2][j+2]) / 27 + nL[i+1][j+1] / 3;
MINMAX3(np1,np2,np3,maxn,minn);
MAX3(max1,max2,maxn,max);
MIN3(min1,min2,minn,min);
max1 = max2; max2 = maxn;
min1 = min2; min2 = minn;
if (max < lab->L[i][j])
max = lab->L[i][j];
if (min > lab->L[i][j])
min = lab->L[i][j];
int val = lab->L[i][j] + params->sharpening.amount * diff / 100;
int newL = CLIP(val);
// applying halo control
if (newL > max)
newL = max + (newL-max) * scale / 10000;
else if (newL<min)
newL = min - (min-newL) * scale / 10000;
lab->L[i][j] = newL;
}
}
}
}
}