Points when adding R, G, B frames.

[Serg1]

Points when adding R, G, B frames.
See the picture.
Perhaps this is due to the algorithm of frame rotation. Or rounding pixel values when rotated.

[han.k]

Hello Serg,

When you add a few frames only (with some rotation) sometimes some pixels are missed due to rounding effects. This will disappear if you add more images or you could check-mark the black spot filter. See screen shot.

The dark spot filter is only available for mosaic and average stacking. I left it out for sigma-clip since sigma-clip only works well if you stack 10 or more frames and then dark spots (or blue or red) should not occur.

If the you combine 10 frames or more and the spots are still there, please let me test your frames. Then maybe a more advanced algorithm should be applied. I could modify the stacking algorithm to calculate on sub-pixel level but that will be very CPU intensive and stacking will take much more time. This is a route I prefer not to take.

black spot filter.png (21.59 KiB) Viewed 7257 times

[Serg1]

Han, this is not working.
See e-mail.
I folded 1 red 1 green and 1 blue frame. Each frame is obtained by adding several frames.

[han.k]

Hello Serg,

Thanks for the images. Blue it the problem. Blue is also rotated compared with red and green. Interesting test case. I will work on a solution.

Regards, Han

[han.k]

Hello Serg,

Problem solved in version 0.9.122. The blue spots where the summation of two pixels at one reference location. If you rotate one image (blue) sometimes two pixels fit in the diagonal length of one reference pixel. I count now number of pixels values added and also added a black spot routine for the left outs.

You can just add the red, green and blue summation images. Same as you send me.

Han

[Serg1]

Han, the red and green channels need to do the same thing as the blue. The dots are in green and red frames.

[han.k]

Hello Serg,

Strange, I tought it was okay. Maybe it was the handling of the pedestal factor for the red/blue pixels. Anyhow I made some more improvements this morning. This will be version 0.9.123 and released in an hour maybe. If I add the three stacked colors you send me, I get now this:

Han

serg_image.jpg (129.3 KiB) Viewed 7231 times

We will get there. These tests are good for making ASTAP better.

Han

[Serg1]

Han, dots are visible if you put a minimum of 300, maximum of 1500.

[Serg1]

Han, the dots are much smaller than they were. Can you add a separate addition option with sub-pixel precision? With the ability to select the number of sub-pixels.

[han.k]

Han, the dots are much smaller than they were. Can you add a separate addition option with sub-pixel precision? With the ability to select the number of sub-pixels.

Hello Serg,

Can you explain more in detail your request?.

At the moment the stacking is pixel based. So the x,y position is calculated and finally rounded and added to the reference image (lowest HFD). Secondly the number of values added is recorded. So if to one pixel a total of 12 values are added, the final result is divided by 12 so an average. In the Sigma clip stacking, the outliers of these 12 values are removed depending on the sigma setting. This is useful to remove satellite tracks and cosmic rays.

If one of the images is rotated some pixels get more values some less or none. This is later detected and if one pixel received non values, this black position get the value from a neighbor pixel value.

What I could do theoretical after the x,y position is calculated, is to spread the value on four pixels. Below a quick sketch. If the pixels are square you can spread the the pixel value on 4 pixels while stacking. If you include the rotation maybe more pixels. It will increase anyhow the stack time. It will be possible to do this mathematical 100% correct and not so difficult. Rotation between image to stack is rare so you could skip that. But each pixel should should be spit in four values and the amount of splitting should be registrated to calculate later the average. See above example with 12. There is no need for the number of sub-pixels, just an option sub-pixel yes or no. I'm only not sure if this will improve the stacking result much. A point source as a star typical is typical already spread over a 2x2 or 3x3 pixels. You could achieve sub-pixel stacking partly by going from bin2x2 to bin1x1 if you have small sensor pixels.

Regards, Han

ccd shifted rotated.png (19.06 KiB) Viewed 7217 times