Plots sometimes in inverse colors

[Toni Šarić]

Why are plots sometimes in inverse colors?
Usually, I mask part of my image with pure black(pixel value 0) and then I solve an image.
The plots (-indx.png) look fine, the first part is a starfield with red and green circles, and the second part is pure black(which I masked).
But sometimes, plots are inverted, the starfield is all white, and red and green circles are on it.
Example in attachment

[Dustin Lang]

Hi,

That's strange!  Would you be able to attach (or upload to nova.astrometry.net and send a link) an example image after you have masked it?  Are you using FITS format or some other image format?

Maybe the background / bias level of your image is quite high, so zero is much lower than the pixel values?  This looks like it's not inverted - the stars are still brighter than the background, but the background is very bright.  Or maybe there's a signed/unsigned integer issue...?

thanks,

dustin

--
You received this message because you are subscribed to the Google Groups "astrometry" group.
To unsubscribe from this group and stop receiving emails from it, send an email to astrometry+...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/astrometry/dd429e03-8c6e-40af-af7c-83c5d349ee49n%40googlegroups.com.

[Toni Šarić]

Hi Dustin, thanks for reply.

I am attaching the image(masked one) in FITS format which I usually use for solving.

When I solve the same image without masking it, the output plots are normal.

Perhaps the problem is with the masked part, because I set a value to zero, and it is too dark.

Cheers

[Dustin Lang]

Hi Toni,

Right, it looks like the background level in your image is ~ 1500, so the masked regions with value 0 are way darker than the sky level.  For FITS images, the astrometry.net plotting code by default sets the black/white range to pixel percentiles: 25th and 95th percentile.  So with a large region set to zero, that's going to throw off the percentiles :)

cheers,

dustin

To view this discussion on the web visit https://groups.google.com/d/msgid/astrometry/41bf803a-78d6-4271-bbe0-04c076f0591dn%40googlegroups.com.

[Toni Šarić]

I see, thanks for explanation.

What is then better for astrometry solving,  to crop out that part of the image or leave it masked?
Would it mess with noise estimate?

Cheers

Toni

[Dustin Lang]

Hi,

That's a good question, it will slightly disturb the noise estimation.  (That is done using a median, so it's fairly robust, but having a lot of 0.0 values will throw it off a bit.)

cheers,

dustin

To view this discussion on the web visit https://groups.google.com/d/msgid/astrometry/d1c15976-812b-42b7-b120-cf00aed6da92n%40googlegroups.com.