Posted by Jason Edlund on September 30, 2009 at 11:12 PM
I was hoping to get some tips on processing night photos. I have gotten into shooting more night stuff and am trying to learn how to make my photos look as good as possible. The big problem is noise, if anyone can give me some tips I'd be very grateful. Ive included a milkyway shot over Mt Robson from this summer its the first real attempt Ive tried. Thanks for the help
canon 50d iso 3200 60sec
Posted by Barry Southon on October 1, 2009 at 6:57 AM
Jason
What an interesting shot, I too have been trying my hand over here in the UK similar subjects can't help with advice but I'm eager to see what is forthcoming in way of tips.
Can't see the noise here...maybe when you print. What I see here is gorgeous! Well done!
Troup
Posted by Don Miller on October 1, 2009 at 3:55 PM
Assuming you're shooting raw and have the camera set optimally, it's often just an issue of different raw converters and noise reduction programs. You definitely don't want to go after every bit of noise.
In the image above you might want to try NR on just the red channel
What aperture were you using? The time seems long. A wide fast prime stopped down to f2, or a wide 2.8 zoom, is usually my starting point. A stop or two makes a big difference.
Posted by Jason Edlund on October 1, 2009 at 5:08 PM
THanks Don will give that a try.
I was using a 10-22 mm lens at 10mm @ f3.5, the widest it goes and had in camera noise reduction on. I did start to get star streaking and if you zoom in I think its noticable, but tried to help it by duplicating the layer, set it to lighten with a small amount of gaus blur, I got that from Floris' tutorial. I also used the median noise reduction in CS4 and the noise reduction under the filter menu. Any helpful tips on works well for these types of night shots as far as processing them would be appreciated.
Posted by Dan Borzynski on October 1, 2009 at 10:51 PM
Don't take this as sarcastic but I'd probably start looking at the astronomy boards if you plan to do this even semiregularly. It is an issue they deal with all the time and probably have a better practical understanding of dealing with noise than we do. There is even dedicated software out there to handle noise.
Here is what I know of to handle this in future images. (Edit: read the last paragraph for your posted image).
First, take a bunch of 1-3 second exposures. By a bunch, I am looking at 30-100 of them. This takes care of noise in individual images. Next, since the camera is steady but the sky is moving, do a panoramic merge of all of the images. This may require doing it is steps due to system memory constraints. The blending should average any remaining noise in the image. Next, play with saturation, contrast and the color controls to bring out the color.
A second technique has the same starting point - lots and lots of images with the camera tripod mounted but needs another piece of software. There are versions that are freewarres (go to the astronomy boards) or for sale. You create a file with a bunch of layers that have been aligned (some of the software does this for you as well) and in the layer blending options of the software, select an "additive" choice. PS has difference and and multiply, but that doesn;t do the same thing. What happens is that with the additive layer blend, very faint color that might only record as 000000000000001, when you add all of these 1 together, actually gets you that isn't black and the color tools in PS produce something worth looking at. The star may blow out, but again, there is smart software for that as well (this is getting a bit bright, stop adding it together),
Which reminds me of a third technique that might help you image. Take you camera out and make a 1 minute exposure (or whatever time a different shot was taken at) at about the same temperature yours was shot at. Put the lenscap on so you get a theoretically pure black exposure. Open this in PS and make NO adjustments. Probably shoot it at 5500k, no color adjustments, manual color balance. Now make that image a layer in your original photo posted above. Blend the layers as "difference" with 100% opacity. What this will do is subtract off any color noise in the black exposure from the color in your original shot. Since noise is random, this won't be perfect but it might help. Ideally, you do this right after the original shot. From the astronomer photographers I have talked to, this works better than in camera noise reduction, especially when combined an original made with multiple blended exposures.
In all of these, an equitorial mount really helps with sharpness, your mountain aside.
Usually color noise is the biggest problem for these, I've found. Noise reduction doesn't seem to be able to smoothly take care of it in some cases. If that's what you're experiencing, try duplicating the base layer, set the blend mode to color, and do a surface blur (gaussian blur leaves artifacts). This will damage some of the colors in the milky way and other nebulae, but a balance can be found.
Or try some of the astro-methods, while good solutions for signal to noise ratios, they'll leave you with issues with your foreground... requiring some tricky blending.
Posted by Don Miller on October 2, 2009 at 11:28 AM
So at f3 5 you're 2.5 stops slower than a 1.4 lens. At f1.4 your shutter would be about 10 seconds. That would eliminate most noise. The crop camera is not ideal in lens or sensor size, but I got good shots with the 60D (the model before the 10D), so it's not hopeless.
I think the astro photo techniques and landscapers probably are probably different enough that the techniques don't overlap as much as they did with slower cameras and film.
First, we want the shot to look good, astrophotogs want it to be accurate. I don't really care if I get every dim star, I just don't want a lots of false stars from noise.
Second, a wide angle landscape shot is wider than what most astrophotogs mean by "wide field". Shooting wide allows for more earth movement before the stars look unsharp.
With the better cameras for night photography I feel the results are professionally good if taken with the best technique:
1) Look for speed by testing the camera at the max real ISO.
2) Turn on dark frame subtraction, if present. Most better cameras have this.
3) Shoot the best shots with a cool camera. As the camera heats up noise increases. Limit live view, perhaps look at photos on a laptop.
4) Fast glass with an appropriate distance setting. 50mm f1.4 is a inexpensive starting point, but too narrow a fov for most uses.
5) Do stars have color as seen from earth? If you decide the sky should essentially be colorless it's pretty easy to deal with color noise.
6) Are very dim stars important in a print? Some can usually be lost i dealing with luminosity noise.
The key seems to be to get close to true black when printing so the stars pop. This requires good contrast without seeing too much noise. In the OP image I would want to shoot at 15 seconds max. That would get rid of most of the red and blue noise. I don't know how good the 50D is with long exposure. The newer full frame cameras are probably better.
Ideal in Canonland is probably a 5DII with a 1.4 lens stopped down to f2. Care must be taken that the lens doesn't focus past infinity.
An equitorial mount is going to blur the foreground. I have one (just a $300 telescope base that can track earth rotation), but I don't use it for still photos. The advantages of lower ISO with a tracking mount is partially offset by long shutter speed.
Posted by David Bostedo on October 2, 2009 at 6:57 PM
Software like PhotoAcute is designed to do the noise reduction via multiple images like in the Cambridge in Colour tutorial above. However, it means taking several shots, which just might average the stars right out of the image as they move.
Discussion
10 repliescanon 50d iso 3200 60sec
What an interesting shot, I too have been trying my hand over here in the UK similar subjects can't help with advice but I'm eager to see what is forthcoming in way of tips.
Troup
In the image above you might want to try NR on just the red channel
What aperture were you using? The time seems long. A wide fast prime stopped down to f2, or a wide 2.8 zoom, is usually my starting point. A stop or two makes a big difference.
I was using a 10-22 mm lens at 10mm @ f3.5, the widest it goes and had in camera noise reduction on. I did start to get star streaking and if you zoom in I think its noticable, but tried to help it by duplicating the layer, set it to lighten with a small amount of gaus blur, I got that from Floris' tutorial. I also used the median noise reduction in CS4 and the noise reduction under the filter menu. Any helpful tips on works well for these types of night shots as far as processing them would be appreciated.
Here is what I know of to handle this in future images. (Edit: read the last paragraph for your posted image).
First, take a bunch of 1-3 second exposures. By a bunch, I am looking at 30-100 of them. This takes care of noise in individual images. Next, since the camera is steady but the sky is moving, do a panoramic merge of all of the images. This may require doing it is steps due to system memory constraints. The blending should average any remaining noise in the image. Next, play with saturation, contrast and the color controls to bring out the color.
A second technique has the same starting point - lots and lots of images with the camera tripod mounted but needs another piece of software. There are versions that are freewarres (go to the astronomy boards) or for sale. You create a file with a bunch of layers that have been aligned (some of the software does this for you as well) and in the layer blending options of the software, select an "additive" choice. PS has difference and and multiply, but that doesn;t do the same thing. What happens is that with the additive layer blend, very faint color that might only record as 000000000000001, when you add all of these 1 together, actually gets you that isn't black and the color tools in PS produce something worth looking at. The star may blow out, but again, there is smart software for that as well (this is getting a bit bright, stop adding it together),
Which reminds me of a third technique that might help you image. Take you camera out and make a 1 minute exposure (or whatever time a different shot was taken at) at about the same temperature yours was shot at. Put the lenscap on so you get a theoretically pure black exposure. Open this in PS and make NO adjustments. Probably shoot it at 5500k, no color adjustments, manual color balance. Now make that image a layer in your original photo posted above. Blend the layers as "difference" with 100% opacity. What this will do is subtract off any color noise in the black exposure from the color in your original shot. Since noise is random, this won't be perfect but it might help. Ideally, you do this right after the original shot. From the astronomer photographers I have talked to, this works better than in camera noise reduction, especially when combined an original made with multiple blended exposures.
In all of these, an equitorial mount really helps with sharpness, your mountain aside.
Yeah, hit the astronomy boards.
Or try some of the astro-methods, while good solutions for signal to noise ratios, they'll leave you with issues with your foreground... requiring some tricky blending.
Your image is beautiful!
I think the astro photo techniques and landscapers probably are probably different enough that the techniques don't overlap as much as they did with slower cameras and film.
First, we want the shot to look good, astrophotogs want it to be accurate. I don't really care if I get every dim star, I just don't want a lots of false stars from noise.
Second, a wide angle landscape shot is wider than what most astrophotogs mean by "wide field". Shooting wide allows for more earth movement before the stars look unsharp.
With the better cameras for night photography I feel the results are professionally good if taken with the best technique:
1) Look for speed by testing the camera at the max real ISO.
2) Turn on dark frame subtraction, if present. Most better cameras have this.
3) Shoot the best shots with a cool camera. As the camera heats up noise increases. Limit live view, perhaps look at photos on a laptop.
4) Fast glass with an appropriate distance setting. 50mm f1.4 is a inexpensive starting point, but too narrow a fov for most uses.
5) Do stars have color as seen from earth? If you decide the sky should essentially be colorless it's pretty easy to deal with color noise.
6) Are very dim stars important in a print? Some can usually be lost i dealing with luminosity noise.
The key seems to be to get close to true black when printing so the stars pop. This requires good contrast without seeing too much noise. In the OP image I would want to shoot at 15 seconds max. That would get rid of most of the red and blue noise. I don't know how good the 50D is with long exposure. The newer full frame cameras are probably better.
Ideal in Canonland is probably a 5DII with a 1.4 lens stopped down to f2. Care must be taken that the lens doesn't focus past infinity.
An equitorial mount is going to blur the foreground. I have one (just a $300 telescope base that can track earth rotation), but I don't use it for still photos. The advantages of lower ISO with a tracking mount is partially offset by long shutter speed.
For chroma noise reduction, I like this technique from Luminous Landscape, which sounds similar to what you're already doing.