Dear experts,
is there any expierence on using hdrgen & falsecolor on nightviews?
The luminance amplitudes of the interesting scene are very low < 1cd/m². And the few sources are very small. I´m running into noise issues, where my proposed metrics are not working: L_min/L_max < threshold, L_min/L_avg < threshold, gradients < thresholds, - not very clear yet.
Camera settings on Nikon D5200 (mirror reflex camera, 18-55 mm, 1:3,5-5,6):
ISO 100, Aperture-priority-automatic-mode, autobracketing
commands:
hdrgen64bit -o testimg.hdr - r cam.rsp DSC-?.jpg
falsecolor -i testimg.hdr -e -s a -log 2 -l cdm2 > test-false.hdr
Question:
Is there any hidden trick to eliminate noise at low luminance levels OR is my approach nonsense at all? - Our common work on Daylight is not aiming at night-artificial light-scenes. Sorry for abusing our fantastic tools.
Bumping up the ISO is one approach, clear. Changing camera another one?
Thank you and best regards
Robert
Hi Robert,
what you could try (it is popular among amateur astronomers, who aim at
even weaker sources with small image-sensors) is to average multiple
image frames to reduce noise. The number of frames that can be combined
is limited only by the stability of the environment. You could e.g. use
pcomb 001.hdr 002.hdr 003.hdr | pcomb -e ‘ro=ri(1)/3; go=gi(1)/3;
bo=bi(1)/3’ > stack.hdr
I am not sure if hdrgen & Co would do if you feed them several images
with identical exposures. Ideally, they would also just average them and
allow you to “de-noise” the images and combine multiple exposures in one
step and provide some pixel registering.
In any case, it would be good to take readings from a good
illuminance-meter in parallel and compare them with the image integrals.
Who knows what is going on in these cameras. Firmware developers have
more party snap-shots in mind than photometry, when addressing low-light
conditions…
Cheers, Lars.
If you give multiple identical exposures to hdrgen, I believe it will average them in, but it doesn’t help all that much in noise reduction. You are much better off increasing the length of your longest exposure time. Your auto-exposure on the camera may not work at very low light levels. You may need to put the camera into manual mode. Increasing the ISO will not help with noise in general, unless you are up against your maximum exposure time in the longest exposure already.
I once took long-exposure sequences, and found my camera sensor to be very noisy. Hot spots and all. This is waaay back with a 3MP Nikon CoolPix. I played around with jpegpixi and was well impressed by the results. https://sourceforge.net/projects/jpegpixi/ It’s no longer maintained, but you might still be able to download and compile it.
Looks cool. If you build HDRs without automatic alignment, there’s a hot-pixel removal that follows in hdrgen and Photosphere.
Maybe a red herring, but there’s an intriguing evaluation of ISO settings for astrophotography here:
https://www.lonelyspeck.com/how-to-find-the-best-iso-for-astrophotography-dynamic-range-and-noise/
I only wish my skies were clear (and dark) enough to try some of these out…
Hi Robert,
People are stacking and averaging frame sequences to reduce noise in astronomy. This may be hundreds or thousands of frames to get reduced to one image. It is important to have well-aligned images. You may find sensor artifacts that would be invisible in regular photographs, and subtracting an (also stacked&averaged) flat will probably be helpful.
Cheers, Lars.