|
05/09/2011. Here's a question for the signal processors amongst us: which produces the superior image, clips comprised of short exposures that freeze atmospheric blurring at the expense of electronic noise, or clips comprised of longer exposures that do not freeze atmospheric turbulance so well but have far less noise? I think the answer is that it depends on so many factors that it's likely to remain a judgement call at least until I acquire a firmer grasp of the variables. First, I shot a 20-second 15fps clip of emerging sunspot region AR1209 using approximately 40ms exposures and zero gain; then I shot a second 20-second 15fps clip of the same region using 4.2ms exposures at 14db gain. (I need to work out what "a stop" means in terms of db in order to keep the net exposure constant, but nevermind that right now. [6db = factor of 2 = 1 stop; so 14 db is about 2.3 stops which suggests that the short-exposure clip was underexposed significantly relative to the long-exposure clip]). The longer exposures produced (as one might expect) smoother tones with less grain, but in this case the shorter exposures captured more detail. Here are the best images I could produce from each clip. In each case, the best post-processing steps were AVIStack - Photoshop for stretching - Registax for wavelet processing - Photoshop for Focus Magic deconvolution. Obviously, there's a lot of room for adjusting the workflow to get better results from either base image. And clouds moved through after the first image, so I had to wait 20 minutes for the second one. I also had to refocus. Like I said, too many variables to effectively control. Try this again when the two images can be captured back to back, without touching any variables except exposure time and amplifier gain.
Lunt LS60THa50DS + barlow 05/10/2011 I tried again to get a grip on how exposure and gain relate to noise with the Chameleon and the Sun. Today I started with 60ms exposures with zero gain and cut the exposure in half, repeatedly, while raising the gain by 6db. I treated all the pictures alike in an effort to see which combination produced the best and worst results. In each case, I let AviStack 2 stack and combine 300 frames and save the result as a FITS file. This I brought into Registax and saved as a 16-bit TIFF with no stretching or other adjustment. Those I brought into Photoshop and normalized using the levels command to bracket highs and lows with a little room on either side. That I brought back into Registax and processed with Wavelets, choosing a relatively mild sharpening configuration. Those I brought into Photoshop where I first applied a despeckle filter and then deconvolved using FocusMagic (I let FM pick its own parameters based on the same active region; it's important to use the despeckle routine first, because otherwise I think FM mistakes grain for detail and declines to apply a sufficiently strong deconvolution). Here are full-scale portions of the same area of each frame, cropped to fit this page and labelled with the exposure and gain used to capture the clip from which it was distilled. The noise number is the noise index measured by Noise Ninja after wavelet processing (I did not apply Noise Ninja, only used it to measure noise).
60ms 0db noise 38
30ms 6db noise 45
15ms 12 db noise 56
7ms, 18db noise 87
Additional images:
2ms 24db noise 108
3.5ms, 18db noise 75
10ms, 12db noise 63
60ms 0db, with much more aggressive wavelet processing, noise 58
Just at a glance, I'd say the sweet spot is somewhere around 12-15db of gain. The shorter exposures freeze seeing better and the noise is not objectionable. I like the tones captured in the longer exposures at very low gain, but a substantial amount of detail seems to be lost. Be careful not to become so enraptured of very short exposures that they are underexposed, because pulling information out of the "shadows" will emphasize noise. Somewhere around 6-10ms with gains of 12-15 db looks good. If seeing is "fast" you might want to push up to 18db and try for 4ms exposures. At second glance, there's a lot to be said for the really clean exposures at 0db of gain and the versatility of the more aggressive processing they license.
|
:: top ::
© 2011, David Cortner