I photographed the recent 21 August eclipse and have a series of very nice looking images, which, if I can properly combine them should yield a very nice image of the solar corona, prominences, and the lunar disk as illuminated by earthshine. The problem is properly combining the images. Each HDR sequence consists of seven exposures with two stops between successive pairs. Combined with the approximately ten stop range of the camera sensor, a Canon 5diii, that's about twenty-four stops of tonal range or about 16 million to 1. Tone mapping this is a challenge for any piece of software.
Perhaps obviously, I'm thinking of trying PWP's HDR/stacking transformation. But, so far, without luck. Any suggestions? I've tried several without satisfaction: HDR Expose 3, Photomatix, PTAssembler, and Aurora HDR. Thanks.
Cheers, Peter
Eclipse HDR
Moderator: jsachs
Re: Eclipse HDR
To use PWP for this purpose, you first have to take calibration photos of a test target and process them the same way you will process your photos. 24 stops is probably overkill and is beyond what the calibration can handle. The procedure is described in the white paper:
http://www.dl-c.com/Temp/downloads/Whit ... Images.pdf
Once you have assembled the images you will still need to apply a final shaping curve to determine how to map the available gray levels to something you can display and print with 256 gray levels. Usually this takes the form of a contrast-enhancing S-shaped curve.
PWP's approach is to try to create a composite image that preserves a wide dynamic range uniformly across the entire image by reducing the overall contrast. Photomatix takes a different approach by blending different parts from different images, depending on which has the best contrast at any given point. For some landscapes the results are pretty good. I tried Aurora HDR on some images recently and found the resultss garish and highly unrealistic, so I have no idea what they are doing.
Again, bear in mind that the final result will have to squeeze the 24 (or however many) stops into 256 gray levels somehow so nothing is going to make that happen without compromises.
You may have better luck using fewer images - maybe 3 - one exposed for the background sky, one exposed for the highlights and one in between.
http://www.dl-c.com/Temp/downloads/Whit ... Images.pdf
Once you have assembled the images you will still need to apply a final shaping curve to determine how to map the available gray levels to something you can display and print with 256 gray levels. Usually this takes the form of a contrast-enhancing S-shaped curve.
PWP's approach is to try to create a composite image that preserves a wide dynamic range uniformly across the entire image by reducing the overall contrast. Photomatix takes a different approach by blending different parts from different images, depending on which has the best contrast at any given point. For some landscapes the results are pretty good. I tried Aurora HDR on some images recently and found the resultss garish and highly unrealistic, so I have no idea what they are doing.
Again, bear in mind that the final result will have to squeeze the 24 (or however many) stops into 256 gray levels somehow so nothing is going to make that happen without compromises.
You may have better luck using fewer images - maybe 3 - one exposed for the background sky, one exposed for the highlights and one in between.
Jonathan Sachs
Digital Light & Color
Digital Light & Color
Re: Eclipse HDR
Why ? What was the problem ?pbandurian wrote: But, so far, without luck.
Cheers, Peter
Photomatix is probably one of the best, if you did not succeed .....
You could try SNS-HDR : https://www.sns-hdr.com/
Bernard
-
pbandurian
- Posts: 15
- Joined: May 11th, 2009, 7:39 pm
Re: Eclipse HDR
Gentlemen:
Thank you for taking the time to respond. I'm sorry I haven't been able to respond in a timely manner. I will respond to each of you in turn.
Let me be somewhat more specific as to what I did. There were seven shots in each HDR stack all at ISO 200 and f/8 (400mm f/5.6 lens with 1.4 teleconverter). The exposures were 1/2000, 1/500, 1/125, 1/30, 1/8, 1/2, 2 seconds. The Sacks PWP method is, ideally, to calibrate each exposure and map all image tones to one much smaller - compressed - tone range. To do so, one photographs a stepped gray scale target with the identical camera and lens settings used for the HDR stack shots. The only adjustment to be made is to choose a single illumination level for all seven shots. Any illumination level could, in principle, be chosen but one that gives an optimal exposure of the middle stepped gray scale target is an obvious choice. This is a bit fiddly and I wouldn't expect anyone but me to bother with it so I'll describe the results. In describing, it helps to look at the sample test exposures of the Kodak Q-13 target on page 6 of the Stacking Images technical article. Observe that, for example, the 2 stop underexposure image is fairly dark but the steps remain pretty clearly visible in the left, brighter half. This is what we would expect. But, this is for only 2 stops underexposure. For 4 stops underexposure, only about a quarter of the lone steps are visible. With 6 stops, there just isn't much left with which to create a tone curve. Unless I'm all wrong in my understanding of the Sachs PWP approach, while beautiful in the purity and technical correctness of the approach, it may not be practical in this application. That is what I suspected when I first wrote and I still suspect my belief is correct. Please correct me if I'm wrong.
With regard to canned tone mapping programs: as I indicated, I have tried a few. In the past I have found Photomatix generally the most satisfactory. Where it fails in this eclipse HDR stack is in the transition from the lunar/solar limb to the photosphere and low corona. In the composite, the lunar limb is too dark. I presume some of that is due to optical light scattering from the photosphere into the dark lunar disk and the tone mapping algorithms over compensating. But the whole point of this procedure is to mask that effect out. Also, the photosphere and low corona are too bright in the final composite. They look washed out. The details of how these canned programs work is, I'm sure, a carefully kept trade secret. But I can speculate. I expect that the programs have algorithms to select regions, densities, and cutoff/transition rates to select what gets included from which exposure and how its tonality is adjusted. However it is done, the lunar/solar limb to chomosphere and lower corona transition magnitude and rate is too great and too abrupt for the algorithms to handle in a manner this human finds satisfactory. Looking at appropriate regions of the appropriate exposure images, I can see I have much more detail than is visible in the final composite.
Thanks for your thoughts and suggestions. I will look at SNS in the next couple of days. Last week I read a very positive review of the new ON1 HDR software and will try that also. Thank goodness for free trials!
Cheers, Peter
Thank you for taking the time to respond. I'm sorry I haven't been able to respond in a timely manner. I will respond to each of you in turn.
Let me be somewhat more specific as to what I did. There were seven shots in each HDR stack all at ISO 200 and f/8 (400mm f/5.6 lens with 1.4 teleconverter). The exposures were 1/2000, 1/500, 1/125, 1/30, 1/8, 1/2, 2 seconds. The Sacks PWP method is, ideally, to calibrate each exposure and map all image tones to one much smaller - compressed - tone range. To do so, one photographs a stepped gray scale target with the identical camera and lens settings used for the HDR stack shots. The only adjustment to be made is to choose a single illumination level for all seven shots. Any illumination level could, in principle, be chosen but one that gives an optimal exposure of the middle stepped gray scale target is an obvious choice. This is a bit fiddly and I wouldn't expect anyone but me to bother with it so I'll describe the results. In describing, it helps to look at the sample test exposures of the Kodak Q-13 target on page 6 of the Stacking Images technical article. Observe that, for example, the 2 stop underexposure image is fairly dark but the steps remain pretty clearly visible in the left, brighter half. This is what we would expect. But, this is for only 2 stops underexposure. For 4 stops underexposure, only about a quarter of the lone steps are visible. With 6 stops, there just isn't much left with which to create a tone curve. Unless I'm all wrong in my understanding of the Sachs PWP approach, while beautiful in the purity and technical correctness of the approach, it may not be practical in this application. That is what I suspected when I first wrote and I still suspect my belief is correct. Please correct me if I'm wrong.
With regard to canned tone mapping programs: as I indicated, I have tried a few. In the past I have found Photomatix generally the most satisfactory. Where it fails in this eclipse HDR stack is in the transition from the lunar/solar limb to the photosphere and low corona. In the composite, the lunar limb is too dark. I presume some of that is due to optical light scattering from the photosphere into the dark lunar disk and the tone mapping algorithms over compensating. But the whole point of this procedure is to mask that effect out. Also, the photosphere and low corona are too bright in the final composite. They look washed out. The details of how these canned programs work is, I'm sure, a carefully kept trade secret. But I can speculate. I expect that the programs have algorithms to select regions, densities, and cutoff/transition rates to select what gets included from which exposure and how its tonality is adjusted. However it is done, the lunar/solar limb to chomosphere and lower corona transition magnitude and rate is too great and too abrupt for the algorithms to handle in a manner this human finds satisfactory. Looking at appropriate regions of the appropriate exposure images, I can see I have much more detail than is visible in the final composite.
Thanks for your thoughts and suggestions. I will look at SNS in the next couple of days. Last week I read a very positive review of the new ON1 HDR software and will try that also. Thank goodness for free trials!
Cheers, Peter