iceman
16-05-2006, 06:45 AM
Hi all.
DP was kind enough to upload his great data from the 26th (http://www.iceinspace.com.au/forum/showthread.php?t=9991) for all of us to have a play with.
I had a quick play yesterday afternoon. My results are below.
To compare with Dave's process, here's a link to Dave's processed version (http://www.iceinspace.com.au/forum/showthread.php?t=9545&highlight=26th).
The data was great.. the image scale is huge! Due to the sheer size of it, it would probably warrant MAP processing on different features of the planet, and merge them all afterwards. If it was my image, i'd probably go back and do just that. Given the image scale, i'd probably agree with Dave's rating of 9/10 seeing. Though there's always room for improvement, so I'd probably rate it 8.5/10.
I started by splitting the BMP frames into separate R/G/B channels, and feeding them separately into registax.
For this exercise, I didn't do MAP processing. I only aligned on 1 feature; a 128px window centered on the blue festoon to the left and above of Io's shadow. I kept Io and the shadow out of the alignment frame due to their speed of orbit.
I used Gradient estimator in registax. I don't really pay any attention to what the % value is set at, as I just drag the slider when it's finished aligning to the point where the curve starts heading up. With this data, because the seeing was so good, and almost all frames were equally good, I left the slider at the far right and took all 500 ranked frames to the "Optimise" stage.
At the "Optimise" stage, I create reference frame (50 frames), and used wavelets 3@10 and 4@15, hit continue, and them optimise.
I stacked the best 110 frames by dragging the vertical slider down, as the graph was a nice curve, low on the left to high on the right.
Now I did something different. Usually at this step, I will do a medium wavelets process, followed by LR deconvolution of each channel and recombine in AstraImage.
But I didn't do that this time. I was reading an article somewhere, and combined with Janoskiss's insight into deconvolution, that said deconvolution should be done before wavelets.
So I saved the final stacked image from each channel, with no wavelet processing.
Into AstraImage, converted each channel to greyscale and did a LR deconvolution (12/2.5). Recombine and adjust R/B planes slightly. Save as TIFF.
Back into Registax with it, straight to wavelets tab (since it's a single image), and did some mild wavelets adjustment. Gamma curve adjust. Save as TIFF.
Into Photoshop with it, curves and levels adjust and slight unsharp mask.
That's it! I also resized down 75% for the second attachment, to see how it looks.
I'm quite pleased with the result of the new processing method. I plan to go back through my data of the 7th May and see what results I can get with similar processing, and compare to my original processed version.
I don't know if it's better or worse than DP's, but it is a bit different. The third attachment is DP's original.
Happy to hear your comments, and I look forward to seeing others efforts on the same data.
Many thanks to DP for letting me play with his data, it's fun to do when the weather is so bad and you don't have tracking anymore! :)
DP was kind enough to upload his great data from the 26th (http://www.iceinspace.com.au/forum/showthread.php?t=9991) for all of us to have a play with.
I had a quick play yesterday afternoon. My results are below.
To compare with Dave's process, here's a link to Dave's processed version (http://www.iceinspace.com.au/forum/showthread.php?t=9545&highlight=26th).
The data was great.. the image scale is huge! Due to the sheer size of it, it would probably warrant MAP processing on different features of the planet, and merge them all afterwards. If it was my image, i'd probably go back and do just that. Given the image scale, i'd probably agree with Dave's rating of 9/10 seeing. Though there's always room for improvement, so I'd probably rate it 8.5/10.
I started by splitting the BMP frames into separate R/G/B channels, and feeding them separately into registax.
For this exercise, I didn't do MAP processing. I only aligned on 1 feature; a 128px window centered on the blue festoon to the left and above of Io's shadow. I kept Io and the shadow out of the alignment frame due to their speed of orbit.
I used Gradient estimator in registax. I don't really pay any attention to what the % value is set at, as I just drag the slider when it's finished aligning to the point where the curve starts heading up. With this data, because the seeing was so good, and almost all frames were equally good, I left the slider at the far right and took all 500 ranked frames to the "Optimise" stage.
At the "Optimise" stage, I create reference frame (50 frames), and used wavelets 3@10 and 4@15, hit continue, and them optimise.
I stacked the best 110 frames by dragging the vertical slider down, as the graph was a nice curve, low on the left to high on the right.
Now I did something different. Usually at this step, I will do a medium wavelets process, followed by LR deconvolution of each channel and recombine in AstraImage.
But I didn't do that this time. I was reading an article somewhere, and combined with Janoskiss's insight into deconvolution, that said deconvolution should be done before wavelets.
So I saved the final stacked image from each channel, with no wavelet processing.
Into AstraImage, converted each channel to greyscale and did a LR deconvolution (12/2.5). Recombine and adjust R/B planes slightly. Save as TIFF.
Back into Registax with it, straight to wavelets tab (since it's a single image), and did some mild wavelets adjustment. Gamma curve adjust. Save as TIFF.
Into Photoshop with it, curves and levels adjust and slight unsharp mask.
That's it! I also resized down 75% for the second attachment, to see how it looks.
I'm quite pleased with the result of the new processing method. I plan to go back through my data of the 7th May and see what results I can get with similar processing, and compare to my original processed version.
I don't know if it's better or worse than DP's, but it is a bit different. The third attachment is DP's original.
Happy to hear your comments, and I look forward to seeing others efforts on the same data.
Many thanks to DP for letting me play with his data, it's fun to do when the weather is so bad and you don't have tracking anymore! :)