liz/rawTherapee
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Improved 8bit thumbnail cache quality for RAWs; see issue #357

Browse Source
This commit is contained in:
Oliver Duis
2010-11-28 19:11:26 +01:00
parent 20c7a87900
commit f57c3ca799
1 changed files with 47 additions and 18 deletions
Download Patch File Download Diff File Expand all files Collapse all files

61
rtengine/rtthumbnail.cc
View File

@@ -920,30 +920,58 @@ bool Thumbnail::writeImage (const Glib::ustring& fname, int format) {
if (format==1 || format==3) {
// to utilize the 8 bit color range of the thumbnail we brighten it and apply gamma correction
int max = 0;
unsigned char* tmpdata = new unsigned char[thumbImg->height*thumbImg->width*3];
int ix = 0,max;
if (gammaCorrected) {
// if it's gamma correct (usually a RAW), we have the problem that there is a lot noise etc. that makes the maximum way too high.
// Strategy is limit a certain percent of pixels so the overall picture quality when scaling to 8 bit is way better
const double BurnOffPct=0.03; // *100 = percent pixels that may be clipped
// Calc the histogram
unsigned int* hist16 = new unsigned int [65536];
memset(hist16,0,sizeof(int)*65536);
for (int row=0; row<thumbImg->height; row++)
for (int col=0; col<thumbImg->width; col++) {
if (thumbImg->r[row][col]>max)
max = thumbImg->r[row][col];
if (thumbImg->g[row][col]>max)
max = thumbImg->g[row][col];
if (thumbImg->b[row][col]>max)
max = thumbImg->b[row][col];
hist16[thumbImg->r[row][col]]++;
hist16[thumbImg->g[row][col]]+=2; // Bayer 2x green correction
hist16[thumbImg->b[row][col]]++;
}
if (max < 16384)
max = 16384;
// Go down till we cut off that many pixels
unsigned long cutoff = thumbImg->height * thumbImg->height * 4 * BurnOffPct;
int max; unsigned long sum=0;
for (max=65535; max>16384 && sum<cutoff; max--) sum+=hist16[max];
delete[] hist16;
scaleForSave = 65535*8192 / max;
unsigned char* tmpdata = new unsigned char[thumbImg->height*thumbImg->width*3];
int ix = 0;
if (gammaCorrected) {
// Correction and gamma to 8 Bit
for (int i=0; i<thumbImg->height; i++)
for (int j=0; j<thumbImg->width; j++) {
tmpdata[ix++] = gammatab[thumbImg->r[i][j]*scaleForSave >> 13];
tmpdata[ix++] = gammatab[thumbImg->g[i][j]*scaleForSave >> 13];
tmpdata[ix++] = gammatab[thumbImg->b[i][j]*scaleForSave >> 13];
tmpdata[ix++] = gammatab[MIN(thumbImg->r[i][j],max) * scaleForSave >> 13];
tmpdata[ix++] = gammatab[MIN(thumbImg->g[i][j],max) * scaleForSave >> 13];
tmpdata[ix++] = gammatab[MIN(thumbImg->b[i][j],max) * scaleForSave >> 13];
}
}
else {
// If it's not gamma corrected (usually a JPG) we take the normal maximum
max=0;
for (int row=0; row<thumbImg->height; row++)
for (int col=0; col<thumbImg->width; col++) {
if (thumbImg->r[row][col]>max) max = thumbImg->r[row][col];
if (thumbImg->g[row][col]>max) max = thumbImg->g[row][col];
if (thumbImg->b[row][col]>max) max = thumbImg->b[row][col];
}
if (max < 16384) max = 16384;
scaleForSave = 65535*8192 / max;
// Correction and gamma to 8 Bit
for (int i=0; i<thumbImg->height; i++)
for (int j=0; j<thumbImg->width; j++) {
tmpdata[ix++] = thumbImg->r[i][j]*scaleForSave >> 21;
@@ -951,6 +979,7 @@ bool Thumbnail::writeImage (const Glib::ustring& fname, int format) {
tmpdata[ix++] = thumbImg->b[i][j]*scaleForSave >> 21;
}
}
if (format==1) {
FILE* f = g_fopen (fname.c_str(), "wb");
if (!f) {
@@ -987,7 +1016,7 @@ bool Thumbnail::writeImage (const Glib::ustring& fname, int format) {
// (machine dependency is not really an issue, since we all run on x86 and having exactly the same file is not a requirement)
cinfo.dct_method = JDCT_FLOAT;
jpeg_set_quality (&cinfo, 85, true);
jpeg_set_quality (&cinfo, 87, true);
jpeg_start_compress(&cinfo, TRUE);
int rowlen = thumbImg->width*3;
while (cinfo.next_scanline < cinfo.image_height) {