August 21, 2017, Farewell Bend State Recreation Area, Oregon
by Stephen D. Blazier
I wanted to see how much of the sun’s corona I captured on that beautiful day, so I combined my deepest exposures during totality. My longest exposures are 1/13 second at ISO 1600 and f/13, using my Nikon D700 with an AF Nikkor ED 180mm f/2.8D IF lens coupled to a Kenko Teleplus PRO 300 DGX 2x AF Teleconverter, for an equivalent focal length of 360mm.
I took 37 of these RAW images. I chose ViewNX 2 to color balance and convert to 16-bit TIFF for further processing. To get the most out of these I needed to take advantage of both the 14-bit pixel depth and the signal to noise improvement from combining multiple frames. I converted the TIFF images to floating point, aligned them using a Fast Fourier Transform algorithm, and then normalized them by offset and scale. I used a sigma rejection to get rid of outlier pixels, and averaged the rest together. I could have summed them, but since I was working in floating point, the result is equivalent.
Those of us used to teasing faint objects out of the noise have experience mapping from a higher dynamic range to the limited display devices available, but in this case, I wanted both bright and faint detail in the result. I chose to learn about High Dynamic Range (HDR) processing.
It was not easy. I find most HDR documentation is either recipe based (e.g., specifying to click certain buttons in a program without explaining what they do, then experiment with a set of parameters), or analyze technical research papers.
I chose a Debevec creation model, a triangular weighting function, and gamma response curve to produce the HDR image. Other creation models produced somewhat similar results, but like most things with HDR, deserve further exploration. To visualize this HDR image on standard devices, I applied Rafal Mantiuk’s operator, as described in “A Perceptual Framework for Contrast Processing of High Dynamic Range Images”, ACM Transactions on Applied Perception 3, 3 (2006), pp. 286-308.
The result is interesting in several respects, but does not match our actual visual experience of the eclipse. We saw an inner solar corona of pure white light. I don’t believe any output device today can match the brilliance of that inner corona. In my processed image, the corona spans across the field of view, but it faded into the twilight-like sky before that from our vantage in Oregon. Moreover, we saw no hint of the Maria on the moon when we were watching.
Thanks for sharing your experiences viewing and recording the moments of totality with us. You went into another realm with complex post processing it seems. Well done. So far I haven't seen any real life amateur imaging that includes the Earth shine, except in a magazine.
I think Stephen explained in a separate email that the bright spot to the lower left was Regulus.