Is there a way to do a dark frame subtraction?

Bill Boehme wrote in post #10702018
The most important thing is to make the "darks" with the same exposure time as the lights. It also helps to have the camera at the same ambient temperature as when making the original images.

so the noise changes enough with exposure time/temperature that it makes that big a difference? What if we subtracted a 15 second dark frame from a 30 second shot? I've never tried so I'm just curious.

I think you missed the part where I said Pentax has a working system that lets the K7 shoot it's dark frame while you keep shooting normal shots. It does it in the "background" somehow. A friend that I landscape with a lot has one, it's a pretty cool trick.

You have a freeware BlackFrame NR that does exactly what you want.

For short exposures it is not worth it, if you want to reduce noise in short exposures you are better taking several shots of the same scene and then stacking them in average mode in PS or any other software.

jacobsen1 wrote in post #10702848
so the noise changes enough with exposure time/temperature that it makes that big a difference? What if we subtracted a 15 second dark frame from a 30 second shot? I've never tried so I'm just curious.

The exposure time is much more significant than temperature. If the exposure times are not the same then only part of the noise would be subtracted or noise would be added if the dark exposure time were longer than the light exposure.

jacobsen1 wrote in post #10702848
I think you missed the part where I said Pentax has a working system that lets the K7 shoot it's dark frame while you keep shooting normal shots. It does it in the "background" somehow. A friend that I landscape with a lot has one, it's a pretty cool trick.

I would buy a bridge in Arizona before I bought that story (unless it is accumulating partial exposures during idle time intervals while the camera is not busy with an actual exposure). That would work until you had an improbable scenario where you were shooting a bunch of exposures under different ISO settings and and progressively longer exposure times that would hog all of the available RAM.

The outermost masked pixels are for a black reference. Then inside them, there's another rectangular field of pixels that don't do into the final image, but are needed when the camera does the Bayer interpolation. Since every single pixel on the sensor is sensitive to one color only, but a pixel in the final image has three colors (RGB), the camera has to get the additional color information by looking at the pixels around the one being processed. If there weren't extra pixels outside the final image, the interpolation algorithm wouldn't have what's required to work with at the edges of the image.

But according to all the information I've seen about this phenomena, the irregularities in output from the 7D's sensor seems to vary columnwise. According to an optronics specialist I know, averaging the output from about 30 pixels high in a column, and then compare that to other columns, where you do the same, seems to give enough data to compensate for at least most of the problem. If that's true (and he usually knows what he's talking about), then at least if the outermost masked region is wide enough, that should give data required to compute the proper compensation.

I'm an astrophotographer, so perhaps I can add my $0.02 to the discussion.

If you want to get rid of intrinsic camera noise, you really need four reference frames. These are a flat field (which allows for correction of vignetting or lens illumination fall off), a defect map (so you know which pixels produce garbage so you can interpolate them after the fact), a bias frame (essentially a dark frame that's taken with the shortest possible exposure) and a reasonably long dark frame. Each reference image has to be as noise-free as you can get it.

Using the bias and dark frame, you can scale the dark to compensate for any exposure. There's a major caveat though - your camera's noise will typically double for every 5deg C (9deg F) increase in temperature - so some sort of thermal correction is usually necessary when using a DSLR. You would normally use the dark and bias to correct for noise in the flat field too.

Now, to determine whether it's necessary to correct the dark frame, you can just put the cap on the lens and shoot a raw frame of - say - 1 minute duration. You can then examine the pixel counts in your favourite image processing package. Next shoot an image where the camera saturates and divide (pixel wise) the second image by the first. This gives you the true dynamic range of your camera after a 1 minute exposure. Let's say it's 100:1. If the dark frame was 30 seconds, you would have got something like 200:1 and if it was 15 seconds the dynamic range would be about 400:1.

On removal of banding - I find that dark and bias correction is absolutely essential in my DSLR and cooled CCD cameras that I used for imaging the night sky. To create dark frames with the minimum of noise I usually average 100 frames of 1 minute each (and of course 100 bias frames.)

If you're interested in suppressing noise in you camera, you might consider purchasing Jerry Lodriguss' e-book "A Guide to Astrophotography with Digital SLR Cameras". Here's a link http://www.astropix.co​m/GADC/GADC.HTM.

Clearest
Bruce, Toronto