Rejecting hot pixels in small sets

[papaf]

Hi all,
I'm facing a problem and I'm unable to get the better of it. I hope someone will chime in with some suggestion.
I work with a b/w camera, and because of very high time constraint, I shoot very few color images and tend to spend my time on luminance. I generally shoot 3 images per color channel.
The problem is that I cannot get rid of some hot pixels, which admittedly are very few, but still... I use dithering, and I can see in the stacked images that they don't overlap. I can literally count three green, three red and three blue hot pixel. Now, correct me if I'm wrong, but shouldnt non overlapping and clearly out of place pixels get obliterated by pretty much every pixel rejection algorithm? Why are they still there, and what could I do?

Thanks!

Fabio

[Philip de Louraille]

I think that you need 4 or more such frames... What options are you using in the ImageIntegration fields?
1) Image Integration
2) Pixel Rejection (1)

[Nocturnal]

Hi Fabio,

the way I would get rid of such hot pixels is to detect them in your dark frame and averaging them out before stacking. Do you have dark frames that match your lights well enough that you can see the same hot pixels?

Filipe is right that sigma rejection algorithms probably need more than 3 frames to work well. You'd say that with 3 values you could detect an outlier but it depends on how much the outlier sticks out and what your sigma value is. You could experiment with a spreadsheet and a few real values from your stack.

[papaf]

Thanks to the both of you.
I'll have to check tonight for the parameters, but I keep them on defaults.
For the darks, I don't have any, since my camera is Sony based and doesn't need them. I thought with 3, which is the minimum, I could get rid of those few pixels.. I guess they're so few I could even go in and take them out myself...

Fabio

[Nocturnal]

Hi Fabio,

my QHY8 doesn't need darks either but I use them for hot pixel detection. I don't stack with them. I wrote a utility to quickly fix hot pixels.

http://www.tungstentech.com/Software/FixFITS/tabid/78/Default.aspx

[Philip de Louraille]

In the ImageIntegration step, you do not want to use the defaults for the Pixel Reject (1)
Use Winsorized Sigma Clipping and scale + zero offset.

In the ImageIntegration subwindow above, use additive+scaling

[papaf]

Thanks again.
Philip, I tried your settings but I'm afraid I simply have too little info here, as the result is the same.
Sander, will your program do something like the Cosmetic correction script in PI?

Fabio

[Nocturnal]

Fabio,

I don't know what that process does My program finds hot pixels with statistical analysis and fixes them by replacing them with the average of its neighbors. With an OSC camera (default) it uses pixels 2 columns away.

[papaf]

But it still needs darks, right? You can't do it without them?

Fabio

[Nocturnal]

my QHY8 doesn't need darks either but I use them for hot pixel detection. I don't stack with them. I wrote a utility to quickly fix hot pixels.

[Juan Conejero]

I wrote a utility to quickly fix hot pixels.

... which could be ported to PI, even as a command-line utility, or even better, with both a CLI and a GUI ...   

[Nocturnal]

Yes but it's working fine and is already available for free so why bother? If I could make a buck I'd give it a go.