Mosaic RANSAC failure

[Neutronman]

Howdy,

I have done several sucessful mosaics with PI in the past -- 5 panels, 6 panels, etc.  But I am currently stumped with my latest 8 panel I am working on.

I am trying to register the final 2 panels and they consistently fail.  I have tried limiting the reference image to a Preview of the overlapping area, and I've tried increasing RANSAC tolerance all the way up, but still fails.

There is adequate overlap and here is the odd thing...  I can register Panel 1 to Panel 3 just fine, but when trying to register Panel 1 to a composite of Panels 7/5/3 it fails.  This totally confuses me since using the small preview limits the Mosaic tool to look at the exact same stars (in Panel 3) used when registering Panel 1 to Panel 3.  What is different?  This is not making sense to me.

At this point I cannot think of how to fix this.  Any suggestions?

(Using PI 1.7, panels are imaged with FSQ/STL1100)

Thanks for any help!

JOhn

PS -- The pause/abort button does not seem to work during the Mosaic process.  I have to enter Task Manager to force PI to shutdown each time to prevent waiting for endless RANSAC failures 

[RBA]

Try, with a higher value for RANSAC tolerance, also increasing the value for regularity and RMS error, and maybe iterations.

Still, I too would like to know why you succeed when doing P1 to P3 but it fails when aligning P1 to the P7-5-3 composite, using the same previews.

[Neutronman]

Thanks for your input, Rogelio.  Yes, I tried cranking RANSAC tolerance all the way up.  I also just tried craking up Regularity, RMS and Iterations... I even restarted my computer first.  No luck.

Have you ever run into a situation like this where a panel refuses to register within a composite, but registers just fine with a single panel?   Very very odd.

I guess my next attempt will be to start over and register in a different order.  If that does not work I guess I'll try Registar (uless anybody else has another suggestion).

Thanks,

JOhn

[RBA]

Have you ever run into a situation like this where a panel refuses to register within a composite, but registers just fine with a single panel?   Very very odd.

Nope. I have run into situations where a pane would simply not register at all, but not this case where it registers fine to another pane, and then, using the very same previews wouldn't register to a group of already registered panes.

I would still be very interested in the possible reasons why it fails if in theory, the areas being examined are exactly the same.

If you already have Registar, you could use it to generate a reference image, then try StarAlignment on that one, so at least you don't break your workflow and still get better results (plus you'd have to make the image 16 bits, etc).

OTOH I figure there's got to be a reason why you can't use the registered pane that comes from aligning it to just pane 3... You might want to try something with that as well...

[Neutronman]

Well, I got it to work by "sneaking up on it".  Since I could register Panel 1 to Panel 3, I simply used the 1+3 composite to register to the 7+5+3 composite (basically just registered panel 3 to itself to join the halves).  Not elegant but it worked.

I am still not sure why PI was having a problem.  The only difference I can think of is that Panel 3 was translated in the 7+5+3 composite, and obviously NOT translated when I simply registered Panel 1 to Panel 3.

Or it could just be the Pixel Harpies crapping on my mosaic 

JOhn