Nice and thanks! I think the only issue for having this be a general solution (and not just having me rework the source a touch - fine for me, but...) is that scale_colors is protected rather than public.
unpack() is affected by camera_wb only in some (very rare) cases (I can not remember the details, it affects very small set of cameras, may be old Kodaks and/or foveon /old ones, before Sigma Merill).
For modern cameras unpack is not affected by color setting (but default for 'use_camera_matrix' for some cameras depends on wb settings)
subtract_black and scale_colors works with image[][4] array, so you need to call raw2image() before subtract_black().
Also, there is single call for raw2image and subtract_black:
int LibRaw::raw2image_ex(int do_subtract_black);
So:
open_file()
unpack();
raw2image_ex(1);
scale_colors();
will put black-subtracted and white-balanced image in image[][4]
Thank you for all the details! I've been using:
RawProcessor.raw2image()
CFA data then from RawProcessor.imgdata.image[r*RAW_S.iwidth+c][RawProcessor.FC(r,c)]
in my code since ... well, a long time ago... to extract the unprocessed data for Bayer-matrix data. (The idea being, I do my own hot pixel / bad pixel mapping -- this is astrophotography work). When non-Bayer, I run:
RawProcessor.dcraw_process();
red data then from RawProcessor.imgdata.image[r*RAW_S.iwidth+c][0]
green from[1], blue from [2]
I poked around in the debugger and docs last night but still couldn't see the effect of use_camera_wb. The docs (http://www.libraw.org/docs/API-notes.html#imgdata_params) say that unpack() is affected by this but no matter what the setting, the output of raw2image() seems the same.
For Bayer matrices, is there a spot where I can pull off the still-CFA encoded data but having run say black level and color scaling? Or, should I just:
unpack()
subtract_black()
scale_colors()
raw2image()
and find CFA data then from RawProcessor.imgdata.image[r*RAW_S.iwidth+c][RawProcessor.FC(r,c)]
'real picture' is more complex than single raw_image array.
For non-bayer images, raw_image is NULL after unpack(), but color3_image or color4_image is not null (color4 image's 4'th component usually zero, but for 4-shot files from Leaf or Pentax it is not zero).
For floating point (DNGs) float_image, float3_image, float4_image pointers are not zero (bayer, 3-component, 4-component data).
Please note, that row pitch may be greater than raw_with*(element size). real pitch is stored in imgdata.rawdata.sizes.raw_pitch and it is always in bytes (divide by 2 for raw_image to get items, divide by 16 for float4_image, etc).
Also, rawdata.*image arrays contains 'black frame' pixels, so use rawdata.sizes: raw_width/height, top_/left_margin and width/height to access image part containing real image data.
the raw2image() call handles all specifics, so look into source for details :)
Followup:
dcraw_make_mem_image() will skip 4'th component from image[][4] array, so if you need to access full data for de-bayer you need to work with either raw_image or image[][4]
no_interpolation=1 will skip bayer interpolation pass, but all other processing to be done (data scaling, white balance, convert to output color space /useless and even wrong step for not interpolated data).
If you want to do your own data processing and use LibRaw only as data decoder (the way we use it in, for example, FastRawViewer) it may be better to access bayer data directly:
rawdata.raw_image contains pointer to bayer data array 'as decoded' (so, no black level subtracted, but unpacked and linearization curve applied).
If you prefer to use imgdata.image[][4] for demosaic you may use
raw2image() to copy raw_image in image[][4] components
and
subtract_black() to subtract black level
I folllowed you instruction, and it works. I have LibRaw-0.17.2\lib\libraw.a
(996KB) and also have several .exe files at LibRaw-0.17.2\bin folder.
When I executed
simple_dcraw.exe -e myfile.raw
I could get myfile.thumb.jpg
So it works properly now. Thank you very much. As you indicated, I think the mistake I did was that I did not put needed compile options like -DLIBRAW_NODLL
-DLIBRAW_NOTHREADS.
1) I've Installed Qt-5.6.2-mingw-492
2) Run Qt/Mingw prompt (Qt 5.6 desktop from Start menu), than bash in this window
3) cd /path/to/LibRaw-0.17.2
4) mingw-32-make -f Makefile.mingw (no need to run ./configure, but it does not change Makefile.mingw, so it does not matter).
several bin/*.exe created after that, so both library and samples compiles OK.
The proper command line for samples creation (from Makefile.mingw):
Thank you for the response.
You may qt-opensource-windows-x86-mingw492-5.6.2.exe. It has mingw32-make.exe.
I installed it and installed MSYS 1.0.11 where I referenced mingw directory.
After setting the paths to mingw and MSYS, in windows 7 command prompt I compiled libraw by
sh ./configure
mingw32-make -f makefile.mingw
I could get libraw.a (994kb), but not sure whether it is properly compiled or
not.
Because the line
#include "libraw/libraw.h"
in simple_dcraw.cpp suggests that it would be better to put simple_dcraw.cpp to
the parent of samples (../samples), I copied simple_dcraw.cpp to the parent of
samples(../sample) and executed
g++ simple_dcraw.cpp -o simple.exe -Ilibraw -Llib -lraw
at there(../samples). So I think "-Ilibraw -Llib -lraw" seems OK.
I've tried to install Qt+mingw package (Qt 5.8.0-mingw), it installed, but no make(.exe) in this package.
MSYS/mingw setup looks made for someone who is already 'in context' and know what to do to install it.
Your error messages looks like no libraw library is linked. This may be because you've specified wrong library or library path:
g++ simple_dcraw.cpp -o simple.exe -Ilibraw -Llib -lraw
I assume you're already in samples folder, so right path is -L../lib, not -Llib
To use .pro files you may need to edit .pro files to specify (or disable) jpeg library and rawspeed library
If you want to get unprocessed RAW values, it is better to use:
- either imgdata.rawdata.raw_image array (complete unprocessed data)
- or use raw2image() call and use imgdata.image[] array (for bayer images, only one component of 4 will be non-zero)
output_color=0 does not disables white balance and auto-bright
Confirmed.
It looks like VS2013 (in my case) changes the order of execution in this line:
imgdata.makernotes.fuji.FujiExpoMidPointShift = ((short)get2()) / fMAX(1.0f, get2());
Replacing it with
float val0 = get2();
float val1 = get2();
imgdata.makernotes.fuji.FujiExpoMidPointShift = ((short)val0) / fMAX(1.0f,val1);
fixes the problem.
Is divide args should be evaluated left to right or not?
CFA does not define the response. The product of CFA transmission spectra, silicon response, and in-camera raw data processing does. pinv is a very useful tool if you need just a matrix. or going to overload a matrix with curves. Conversion from camera RGB to XYZ is not useful however. There is no need in intermediate steps.
I think I figured it out: it makes physically more sense to express the XYZ primaries in terms of the primaries of the CFA. The best you can then do in the colors from the CFA to XYZ, is to minimize the error made. This is what is implemented in the pseudoinverse() function, as described here: http://www.sci.utah.edu/~gerig/CS6640-F2012/Materials/pseudoinverse-cis61009sl10.pdf
This is done by minimizing the squared error of an overconditioned system of equations.
I'm not excellent in color theory, but in general:
- camera responds to any visible spectrum signal
- and, also, on UV and IR wavelenghts too.
So, triangle angles (bayer primaries) may reside outside of 'human eye locus' on xy 'visible colors' diagram (to include as much as possible colors into triangle).
The values you calculate from camera color profile data to be derived from profile creator's intention (color gamut, etc), not real 'bayer primaries'.
Hi,
Thanks for the quick response!
I tried to use "output_color = 0" and "half_size = 1" at the same time, the output looks like a channel is missing.
Is there a way that I can have a unprocessed raw in half size? Perhaps I am unaware of something?
Thanks a lot,
We assume that someone who uses 'parts' of dcraw_process() will subclass LibRaw :)
Nice and thanks! I think the only issue for having this be a general solution (and not just having me rework the source a touch - fine for me, but...) is that scale_colors is protected rather than public.
unpack() is affected by camera_wb only in some (very rare) cases (I can not remember the details, it affects very small set of cameras, may be old Kodaks and/or foveon /old ones, before Sigma Merill).
For modern cameras unpack is not affected by color setting (but default for 'use_camera_matrix' for some cameras depends on wb settings)
subtract_black and scale_colors works with image[][4] array, so you need to call raw2image() before subtract_black().
Also, there is single call for raw2image and subtract_black:
int LibRaw::raw2image_ex(int do_subtract_black);
So:
open_file()
unpack();
raw2image_ex(1);
scale_colors();
will put black-subtracted and white-balanced image in image[][4]
Alex,
Thank you for all the details! I've been using:
RawProcessor.raw2image()
CFA data then from RawProcessor.imgdata.image[r*RAW_S.iwidth+c][RawProcessor.FC(r,c)]
in my code since ... well, a long time ago... to extract the unprocessed data for Bayer-matrix data. (The idea being, I do my own hot pixel / bad pixel mapping -- this is astrophotography work). When non-Bayer, I run:
RawProcessor.dcraw_process();
red data then from RawProcessor.imgdata.image[r*RAW_S.iwidth+c][0]
green from[1], blue from [2]
I poked around in the debugger and docs last night but still couldn't see the effect of use_camera_wb. The docs (http://www.libraw.org/docs/API-notes.html#imgdata_params) say that unpack() is affected by this but no matter what the setting, the output of raw2image() seems the same.
For Bayer matrices, is there a spot where I can pull off the still-CFA encoded data but having run say black level and color scaling? Or, should I just:
unpack()
subtract_black()
scale_colors()
raw2image()
and find CFA data then from RawProcessor.imgdata.image[r*RAW_S.iwidth+c][RawProcessor.FC(r,c)]
Craig
And second followup (several notes):
'real picture' is more complex than single raw_image array.
For non-bayer images, raw_image is NULL after unpack(), but color3_image or color4_image is not null (color4 image's 4'th component usually zero, but for 4-shot files from Leaf or Pentax it is not zero).
For floating point (DNGs) float_image, float3_image, float4_image pointers are not zero (bayer, 3-component, 4-component data).
Please note, that row pitch may be greater than raw_with*(element size). real pitch is stored in imgdata.rawdata.sizes.raw_pitch and it is always in bytes (divide by 2 for raw_image to get items, divide by 16 for float4_image, etc).
Also, rawdata.*image arrays contains 'black frame' pixels, so use rawdata.sizes: raw_width/height, top_/left_margin and width/height to access image part containing real image data.
the raw2image() call handles all specifics, so look into source for details :)
Followup:
dcraw_make_mem_image() will skip 4'th component from image[][4] array, so if you need to access full data for de-bayer you need to work with either raw_image or image[][4]
no_interpolation=1 will skip bayer interpolation pass, but all other processing to be done (data scaling, white balance, convert to output color space /useless and even wrong step for not interpolated data).
If you want to do your own data processing and use LibRaw only as data decoder (the way we use it in, for example, FastRawViewer) it may be better to access bayer data directly:
rawdata.raw_image contains pointer to bayer data array 'as decoded' (so, no black level subtracted, but unpacked and linearization curve applied).
If you prefer to use imgdata.image[][4] for demosaic you may use
raw2image() to copy raw_image in image[][4] components
and
subtract_black() to subtract black level
instead of dcraw_process()
I folllowed you instruction, and it works. I have LibRaw-0.17.2\lib\libraw.a
(996KB) and also have several .exe files at LibRaw-0.17.2\bin folder.
When I executed
simple_dcraw.exe -e myfile.raw
I could get myfile.thumb.jpg
So it works properly now. Thank you very much. As you indicated, I think the mistake I did was that I did not put needed compile options like -DLIBRAW_NODLL
-DLIBRAW_NOTHREADS.
ST
1) I've Installed Qt-5.6.2-mingw-492
2) Run Qt/Mingw prompt (Qt 5.6 desktop from Start menu), than bash in this window
3) cd /path/to/LibRaw-0.17.2
4) mingw-32-make -f Makefile.mingw (no need to run ./configure, but it does not change Makefile.mingw, so it does not matter).
several bin/*.exe created after that, so both library and samples compiles OK.
The proper command line for samples creation (from Makefile.mingw):
g++ -O4 -I. -w -DLIBRAW_NODLL -DLIBRAW_NOTHREADS -o bin/raw-identify samples/raw-identify.cpp -L./lib -lraw -lws2_32 -lm
(or change to -L../lib if you're compiling within samples folder).
Thank you for the response.
You may qt-opensource-windows-x86-mingw492-5.6.2.exe. It has mingw32-make.exe.
I installed it and installed MSYS 1.0.11 where I referenced mingw directory.
After setting the paths to mingw and MSYS, in windows 7 command prompt I compiled libraw by
sh ./configure
mingw32-make -f makefile.mingw
I could get libraw.a (994kb), but not sure whether it is properly compiled or
not.
Because the line
#include "libraw/libraw.h"
in simple_dcraw.cpp suggests that it would be better to put simple_dcraw.cpp to
the parent of samples (../samples), I copied simple_dcraw.cpp to the parent of
samples(../sample) and executed
g++ simple_dcraw.cpp -o simple.exe -Ilibraw -Llib -lraw
at there(../samples). So I think "-Ilibraw -Llib -lraw" seems OK.
Sorry, could not help
I've tried to install Qt+mingw package (Qt 5.8.0-mingw), it installed, but no make(.exe) in this package.
MSYS/mingw setup looks made for someone who is already 'in context' and know what to do to install it.
Your error messages looks like no libraw library is linked. This may be because you've specified wrong library or library path:
g++ simple_dcraw.cpp -o simple.exe -Ilibraw -Llib -lraw
I assume you're already in samples folder, so right path is -L../lib, not -Llib
To use .pro files you may need to edit .pro files to specify (or disable) jpeg library and rawspeed library
You'll get white balance (daylight, because of use_camera_wb=0), demosaic and output range scaling.
Thanks for the quick reply!
With the following setting:
OUT.use_camera_matrix = 0;
OUT.output_color = 0;
OUT.highlight = 0;
OUT.use_camera_wb = 0;
OUT.gamm[0] = 1.0;
OUT.gamm[1] = 1.0;
OUT.no_auto_bright = 1;
Will I get a demosaic raw without any post-processing?
If you want to get unprocessed RAW values, it is better to use:
- either imgdata.rawdata.raw_image array (complete unprocessed data)
- or use raw2image() call and use imgdata.image[] array (for bayer images, only one component of 4 will be non-zero)
output_color=0 does not disables white balance and auto-bright
the patch above contains typo (missing bracket)
latest one correct this: https://github.com/LibRaw/LibRaw/commit/e818b3a0f75c6c4d1c18a2b003b54e25...
Hi Alex,
Thanks for the quick fix!
Use this patch: https://github.com/LibRaw/LibRaw/commit/b622cddaebfb749cad42b66f03838592...
Fortunately, it looks like this is only single problem point, no more 'two getN() divided in one statement' (sum and multiply will work OK)
Confirmed.
It looks like VS2013 (in my case) changes the order of execution in this line:
imgdata.makernotes.fuji.FujiExpoMidPointShift = ((short)get2()) / fMAX(1.0f, get2());
Replacing it with
float val0 = get2();
float val1 = get2();
imgdata.makernotes.fuji.FujiExpoMidPointShift = ((short)val0) / fMAX(1.0f,val1);
fixes the problem.
Is divide args should be evaluated left to right or not?
CFA does not define the response. The product of CFA transmission spectra, silicon response, and in-camera raw data processing does. pinv is a very useful tool if you need just a matrix. or going to overload a matrix with curves. Conversion from camera RGB to XYZ is not useful however. There is no need in intermediate steps.
I think I figured it out: it makes physically more sense to express the XYZ primaries in terms of the primaries of the CFA. The best you can then do in the colors from the CFA to XYZ, is to minimize the error made. This is what is implemented in the pseudoinverse() function, as described here:
http://www.sci.utah.edu/~gerig/CS6640-F2012/Materials/pseudoinverse-cis61009sl10.pdf
This is done by minimizing the squared error of an overconditioned system of equations.
I'm not excellent in color theory, but in general:
- camera responds to any visible spectrum signal
- and, also, on UV and IR wavelenghts too.
So, triangle angles (bayer primaries) may reside outside of 'human eye locus' on xy 'visible colors' diagram (to include as much as possible colors into triangle).
The values you calculate from camera color profile data to be derived from profile creator's intention (color gamut, etc), not real 'bayer primaries'.
if you do not want any color profile, you need to set params.output_color=0
With this setting image to be white balanced, demosaiced and scaled to fill 16-bit range.
So calling dcraw_process() with just the settings written above will create a RGB image without any further kind of processing? Thanks!
noise reduction and highlight recovery are not enabled by default
Pages