|
|
Untitled_blackout.png (700,323 bytes) |
|
|
Option 2 has my preference for the following reasons: Focussing: when not suppressing double reflections the average HFD will depend on the average of the direct star image and its reflection, but as the reflection has its best focus in another plane (while the direct image of the star reduces in size its reflection will get larger), the software will end up with a focus distance which is not at the slit, causing lower efficiency in the spectroscope. Of course it will help when focussing is done on the brightest star alone. Plate-solving: the double reflections add 'new' stars to the image, which in turn are detected by the plate-solve software. This can be mitigated by lowering the exposure time, but that will cause issues when doing a slew with an offset to get a very bright star on the slit. With that procedure three slews, two plate-solves, and one focussing procedure are involved. As we cannot predict what the star field will look like it must assume that three different exposure times are required. Masking the double reflection will avoid this issue. We have seen that brute-force method works well (thanks Han!), but it could be improved by not looking at each bright pixel and adding a mask for that, but to find the bright centroid, calculate its HFD, and generate the mask based on that, in which the diameter of the masking dots is a percentage of the HFD (generally a value larger than 100% would be required). It is no problem if that takes an additional second or two, as this masking is only required during the initial stage of the spectroscopic sequence (and during focussing). Nicolàs |
|
|
I agree it is good to have your masking method with option 2, this reflections are very annoying so having an option to get rid of them is interesting. I never try the mask personally, but the good experience by Nicolàs is important. Personally I autofocus only once per night and I select a field without bright star for that. This probably work for me because I have a RC with a carbon tube. It need to take account for the slit orientation in the camera, it can be horizontal as in Nicolàs image, but also vertical, other angle are not used. It also need parameters for the reflection distance from the star and diameter of the reflection. One reflection is always larger than the other but it is probably sufficient to have the diameter of the larger one. This also need an option to deactivate the function because other spectroscope use a polished metallic slit that not have secondary reflection. |
|
|
The way Han has currently implemented it is by defining both dX and dY for the centre of the masking dots relative to the direct image of the star. If I am right the dX is positive for the upper reflection and negative for the lower, so that it can still work with a slit with a non perpendicular orientation. In an ideal world the double reflections could be automatically detecting by aiming the telescope at a very bright star (only once to detect the offsets, which can then be stored, so similar to doing a guiding calibration run). Currently I have these masking dots in my guider/finder darks. |
|
|
I lost my code {rare but it happened} but here is a new version. It is a very simple brute force approach to blackout the two reflections. Threshold default something like 5000. dx=0, dy=45, diameter=10. No star detection. Just simply a threshold. Any pixel above the threshold results in two round black spot drawn at dx, dy distance. corrected code: procedure TFits.ghost_blackout(threshold,dx,dy,diameter: integer);//blackout of ghost stars for spectroscopy var jy,jx,m,n,xm,ym,qy,rx, radius,sqrradius,ny,nx: integer; procedure setpixel(ay,ax : integer; val: single); begin if ((ax>=0) and (ax<Fwidth) and (ay>=0) and (ay<Fheight)) then Fimage[0,ay ,ax]:=val; end; function getpixel(ay,ax: integer): single; begin if ((ax>=0) and (ax<Fwidth) and (ay>=0) and (ay<Fheight)) then result:=Fimage[0,ay ,ax] else result:=9999999; end; begin Radius:=diameter div 2; sqrradius:=sqr(diameter) div 4; try for jy:=1 to Fheight-2 do for jx:=1 to Fwidth-2 do begin if ((fimage[0,jy , jx ]>threshold) and (fimage[0,jy+1 ,jx ]>threshold) and (fimage[0,jy ,jx+1]>threshold) and (fimage[0,jy+1 ,jx+1]>threshold)) then begin if ((getpixel(jy+dy,jx+dx))< (getpixel(jy,jx))) then //this is a fainter ghost for ny:=-radius to +radius do for nx:=-radius to +radius do if sqr(ny)+sqr(nx)<=sqr(radius) then setpixel(jy+ny+dy,jx+nx+dx,$000000); //blackout ghost above if ((getpixel(jy-dy,jx-dx))< (getpixel(jy,jx))) then //this is a fainter ghost for ny:=-radius to +radius do for nx:=-radius to +radius do if sqr(ny)+sqr(nx)<=sqr(radius) then setpixel(jy+ny-dy,jx+nx-dx,$000000); //blackout ghost below end; end; except end; end; Untitled.png (1,088,532 bytes) |
|
|
Thanks Han, that brute-force method should do. Looking forward testing it! :-) Nicolàs |
|
|
I have updated the code with a centering routine. This should work reliable. The only problem is how to squeeze the four input parameters in the spectroscopy tab which is already full. But I leave that to Patrick. procedure TFits.ghost_blackout(threshold,dx,dy,diameter: integer);//blackout of ghost stars for spectroscopy var jy,jx,radius,sqrradius,ny,nx,fwidth,fheight,cx,cy,fluxmain,fluxghost1,fluxghost2,gy,gx: integer; procedure setpixel(ay,ax : integer; val: single); begin if ((ax>=0) and (ax<Fwidth) and (ay>=0) and (ay<Fheight)) then fimage[0,ay ,ax]:=val; end; function getpixel(ay,ax: integer): single; begin if ((ax>=0) and (ax<Fwidth) and (ay>=0) and (ay<Fheight)) then result:=fimage[0,ay ,ax] else result:=0; end; function center_of_gravity(iy,ix : integer; out cy,cx,flux : integer): boolean; var sumval,sumvalx,sumvaly : double; val : single; i,j,signal_counter: integer; begin result:=false;//default failure flux:=0;//default failure SumVal:=0; SumValX:=0; SumValY:=0; signal_counter:=0; for j:=-diameter to diameter do //search wide enough to center on donuts for i:=-diameter to diameter do begin val:=getpixel(iy+j,ix+i); if val>threshold div 2 then begin SumVal:=SumVal+val; SumValX:=SumValX+val*(i); SumValY:=SumValY+val*(j); inc(signal_counter); {how many pixels are illuminated} end; end; if signal_counter<4 then //no star detected begin cy:=iy; cx:=ix; end else begin //center of gravity cy:=round(iy+SumValY/SumVal); cx:=round(ix+SumValX/SumVal); flux:=round(Sumval); result:=true; end; end; begin Radius:=diameter div 2; sqrradius:=sqr(radius); fheight:=length(fimage[0]); fwidth:=length(fimage[0,0]); sqrradius:=sqr(diameter) div 4; try for jy:=1 to Fheight-2 do for jx:=1 to Fwidth-2 do begin if ((fimage[0,jy , jx ]>threshold) and (fimage[0,jy+1 ,jx ]>threshold) and (fimage[0,jy ,jx+1]>threshold) and (fimage[0,jy+1 ,jx+1]>threshold)) then //four pixels above threshold begin if center_of_gravity(jy,jx,cy,cx,fluxmain) then //found a possible true star begin if getpixel(cy+dy,cx+dx)<>0 then center_of_gravity(cy+dy,cx+dy,gy,gx,fluxghost1) //calculate ghost1 flux else //no need to check fluxghost1:=0; if getpixel(cy-dy,cx-dx)<>0 then center_of_gravity(cy-dy,cx-dy,gy,gx,fluxghost2) //calculate ghost2 flux else //no need to check fluxghost2:=0; if ((fluxghost1<fluxmain) and (fluxghost2<fluxmain)) then //found the main star and not a ghost begin if getpixel(cy+dy,cx+dx)<>0 then //ghost star is not yet blackouted for ny:=-radius to +radius do for nx:=-radius to +radius do if sqr(ny)+sqr(nx)<=sqrradius then //round spot setpixel(cy+dy+ny,cx+dx+nx,$000000); //blackout ghost1 above if getpixel(cy-dy,cx-dx)<>0 then //ghost star is not yet blackouted for ny:=-radius to +radius do for nx:=-radius to +radius do if sqr(ny)+sqr(nx)<=sqrradius then //round spot setpixel(cy-dy+ny,cx-dx+nx,$000000); //blackout ghost2 below end; end; end; end; except end; end; ghost_blackout v2.jpg (322,907 bytes) |
|
|
Thank you Han, I add your method to mask this ghost reflection. The new method is called after an eventual dark removal and noise filter. The settings are in the internal guider camera tab. But to make it not intrusive for people that not need it, this is visible only if the spectroscopy function are active, and the detail setting is visible only if the main option to remove the reflection is checked. There is no additional setting for the finder, but if the finder camera is the same as the guider the ghost reflection are also removed using the same settings. This is available in version 0.9.94-4088 : https://vega.ap-i.net/pub/ccdciel/daily_build/ |
|
|
Thanks Patrick and Han, will test it as soon as weather allows. |
|
|
I can now confirm that the masking algorithm works fine, once again thanks Han and Patrick! |
|
|
Thank you for testing Nicolàs. I make a change today so the mask position and size are adjusted automatically for binning change. What it do is to save the current binning when you adjust the position. If the binning change later the x,y,size number in the interface do not change but this is adjusted in the ghost_blackout() procedure using the binning in the current file header. This way it work when using a different binning for the autofocus or the finder. This can be tested with version 0.9.94-4093 : https://vega.ap-i.net/pub/ccdciel/daily_build/ |
|
|
I have added some lines in the documentation about this new feature, tab camera and tab spectroscopy. |
|
|
Thanks! |
|
|
Thank you for the documentation Han. |
- Anonymous
