28 Feb
2009
28 Feb
'09
8:23 a.m.
I gave Patrick’s 2nd batch of the HorseHead data a shot.
http://www.slas.us/gallery2/main.php?g2_itemId=1173
ONLY READ ON IF you are interested in MaxIM DL and PhotoShop image processing.
I will try to give full disclosure here. First of all, obviously Tyler is the target; in other words he has set the standard. So I downloaded a copy of his HH image processed data as contained in Patrick’s (his data) gallery site: http://www.slas.us/gallery2/main.php?g2_itemId=1025 . I used that image as a model just to look at and see if I could come close to it with Patrick’s 2nd batch of data.
I used Joe’s way of processing as follows.
1 - I took each of the clear photos and subtracted the dark from it.
2 - Then I used a "remove bad pixel" feature of MaxIm DL to get rid of bad pixels, some of which were in more
than one of the frames. The feature keeps a map of bad pixels so you can apply it
to other views from the same session.
3 - Next I picked out the clear that seemed to show the best range and used an "equalize screen stretch" button to
make all the clears about the same.
4 - I combined them using various methods to see which was the best. I think the best one turned out to be
taking a medium of the images.
5 - I did the same sort of thing for the colors, except that one -- I think it was green -- had only one image so I didn't combine it with another green.
6 - Finally I Color Combined LRBG without playing with curves because it seemed about right as it was. My last operation was to adjust the
brightness and contrast with MaxIm DL (Max).
<End Joe’s Way>
I did 3 additional things in Max before saving the image as a TIFF. If the image is saved in Max at that point and brought into PhotoShop (PS) then you have a black image. The image has to be stretched using curves and levels to see anything because of the way Max and PS approach things differently. So, I used Max to stretch the data first because when comparing the two I felt Max did a better job for the initial stretch than PS. Here is what I did.
(I am going into detail because this is a learning thing. Also if there is a better way, maybe Tyler can set us straight.)
1 – In the Screen Stretch box I manually stretched the right had carrot as far to the right as I could without clipping any star data.
2 – In Max I selected menu item Process > Stretch and set these parameters Linear Only, Screen Stretch, and 16-bit and clicked OK.
3 – Again in Max I selected menu item Process > Stretch again and set these parameters Gamma (0.5), Max Pixel, and 16-bit and clicked OK.
Then I saved the image as a TIFF for Photoshop.
>From here on it gets fuzzy because I did a lot of playing around.
In PS the image will still be a little dark. So Using Tyler’s image as a guide I used curves and levels in the RGB to bring up the contrast and brightness in moderate steps keeping an eye on the histogram to be sure I wasn’t clipping data. When you see the combing effect in the histogram you know you are stretching the data pretty far. The noise will also increase as you boost the brightness and contrast. When I got the image to about what Tyler’s brightness looked like then I went into the Blue level and just stretched it a bit to give the background a bit more of a blue hue than straight black.
Now I checked Tyler’s image and noticed his stars were smaller than mine. Size matters, so I created a mask for just the big stars by :
1. Creating a duplicate image and doing a Hi-pass filter with threshold = 1
2. Did a Gaussian blur with threshold = 1
3. Went into Image > Adjust > Threshold and played with the slider till mostly the big stars were left. I cloned out anything I didn’t want till I had the mask I wanted.
4. Then I did a menu Select >All, then Edit > Copy.
5. In the original image I created a duplicate layer of the original image, then went to the channels tab and created a new channel. I pasted the mask into the new color channel. I then clicked on the eye of the RGB channel to reveal the image, then went back to the adjustment layer and clicked on it. From there I went to the menu items. I clicked on Select > Load Selection and lo-and-behold all my big stars were selected.
6. With the big stars selected I enlarged the image to get a closer look at the selection. The selection wasn’t big enough so I went to menu item Select > Mofiy > Expand and increased the size of the selection by 1 pixel.
7. Satisfied with the selection I went to menu Filter > Other > Minimum and it shrank the stars.
If you look at some of the stars you will see a halo around them. I don’t have this process down perfect yet.
I felt my image was looking pretty good at this point but Tyler’s image was much smoother. Being this close I had to go all the way, so I went out on the web and bought NeatImage as a plug-in to PS. Once I applied NeatImage I felt it smoothed out the noise very well and can give Tyler a run for his money. That may have sounded boastful but it was meant as a thank you for helping us overcome the huge learning curve in image processing. And no less thanks to Patrick and Joe and all of you who have added your input to help us fledglings along.
Jim