I am trying to use raw2hdr to generate HDRI of the face of a light source, and I keep getting pink noise overlaying structures in the background. I am not getting any errors during the process. Does anyone know why this might be happening, and if there’s a way around this while still using the raw images? I put the jpegs through photosphere, and the pink noise doesn’t appear, so it seems to be a calculation error in the combination of the raw LDR images.
I work in research, so I can’t post the entire image, but here is a crop of the background to give an idea of what I am seeing.
This may be caused by an issue with dcraw where it doesn’t leave quite enough headroom in white. Try adding “-b 1.2” to the raw2hdr command line, and increase it further if you need to by increments of 0.1. (The default brightness value is 1.1)
I went through and tested different values, but the pink did not go away. Instead, there was in increase in blue noise. @Greg_Ward do you have any other suggestions?
At this point I would have to see your input images to figure out what is going on. You can contact me off-list (gward@lmi.net) and I’ll see what I can find. If you’re worried about confidentiality, then take photos of something less sensitive with the same problem and I’ll work with those.
Hi Greg and Lia,
I also encountered identical problem when I generated a HDR image using several raw files.
Here’s my code:
raw2hdr -h -w -o output.hdr input1.CR2 input2.CR2 … input15.CR2
The running process was without any errors and warnings, but the rendering was produced with higher pink noises. I know that the “raw2hdr” program is executed based on the “dcraw”, “hdrgen”, and “exiftool” programs. All the programs were derived from the latest released source code. I have also tried some parameters (e.g., -a -e -g -f) in the code line to see how the results would be different, but I still got the similar results.
I don’t know whether you solved the problem or not. I will appreciate any help.
I did try different settings for the -b option (i.e., default, 2, 2.2, 2.3, 6, and 8), but the problem was not solved as expected. For example, the pink noises were decreased with the -b option being set from default to 2.2, but they became higher again with the -b option being set to 2.3. The values of 6 and 8 would make the images totally white. It seems that such an option cannot effectively improve the quality of the image. I will attach the results for your reference.
Some information derived from the process is shown as below:
#########
Loading Canon EOS 80D image from input1.CR2 …
Scaling with darkness 0, saturation 16383, and
multipliers 1.832031 1.000000 1.707031 1.000000
Building histograms…
Writing data to standard output …
1 image files updated
#########
Whether the values of the saturation and the multiplier would affect the results?
(-b default)
(-b 2)
