thumb nail images (nn resampling)>
thumb nail images (bilinear interpolation resampling)>
Last Updated: April 4, 1996
M. Robinson
Starry, starry night...
aspect correct comet>
Well there it is aspect corrected and now sort of calibrated (see calibration Notes below). I did the aspect thing by using the idl function congrid
fixed_image=congrid(comet,537,414,/interp)
I got the number of lines by (161/95)*244 = 414. The INTERP keyword signifies a linear interpolation. This image is an average of calibrated versions of 30870000 and 30870040 (both clear filter 999 msec). In the registration 30870000 was adjusted to match the geometry of 30870040. You may notice a little funkiness around the edges - this "shows off" the offset in registering the two files. If you download this file and stretch it a bit harder you can imagine the comet shape a little better. If you want the unstreched data (32 bit format) let me - I'll ftp it to you (for this file or all the calibrated versions). I am working on a cleaned up pretty picture from this and it should be available soon. I must say it is interesting working with data 10-20 dns above background (especially when the dark has noise +- 5 dn!).
This is a before aspect correction version of the same file shown above.
before aspect correction comet>
Below are two thumbnail pages for reference - images subsampled by 4x. They are identical except for the resampling scheme. Mike is concerned that team members will be "totally mislead" by the original thumbnail due to the obvious artifacts introduced by nearest neighbor subsampling, he is afraid that team members will only look at the thumbnail and not the actual images. To avoid any confusion I have made a second version that has used a bilinear interpolation scheme -------> Additionally, I urge all team members to look at the actual images and not rely on the thumbnail for anything other than it was intended, a Q&D guide for your efforts! I certainly apologize if I "totally mislead" anyone ;-)! Enjoy!!!
thumb nail images (nn resampling)>
thumb nail images (bilinear interpolation resampling)>
Compare this to table Jeff Warren made from Jimbos web page shown down below.
File Name Temp EXP Filt Comment --------------------------------------------------------------------- 30850007.FIT -30.1 999 7 Missing lines, Bogus dns? 30860007.FIT -30.1 999 7 dark frame 30870000.FIT -30.1 999 0 "Best" Comet image 30870010.FIT -30.1 250 0 yawn 30870021.FIT -30.1 999 1 ... 30870032.FIT -30.1 999 2 ... 30870040.FIT -30.1 999 0 "Best" compressed comet image 30870050.FIT -30.1 500 0 pretty dim 30870065.FIT -30.1 999 5 ... 30870071.FIT -30.1 999 1 noise pattern switch in image 30870082.FIT -30.1 999 2 a bit dim 30870095.FIT -30.1 999 5 a bit dim 30870100.FIT -30.1 999 0 NW position - no comet? 30870110.FIT -30.1 999 0 NW position - no comet? 30870120.FIT -30.1 999 0 bad data 30870130.FIT -30.1 999 0 NE position, comet visible 30870140.FIT -30.1 999 0 SE comet visible 30870150.FIT -30.1 999 0 SW position - no comet? 30870160.FIT -29.6 999 0 SW position - no comet? 30870170.FIT -29.6 999 0 From here down are the tail images.... 30870180.FIT -30.1 999 0 Interesting temp jump here 30870190.FIT -29.6 999 0 Wild tiger striping starts here 30870200.FIT -29.6 999 0 30870210.FIT -29.6 999 0 30870220.FIT -29.6 999 0 30870230.FIT -29.6 999 0 30870240.FIT -29.6 999 0 End tail mosaic 30870257.FIT -29.6 10 7 dark frame 30870267.FIT -29.6 999 7 dark frame Notes: EXP = exposure time in milliseconds, Temp C, Filt = Filter Code 7=1050nm, 0=clear, 1=550nm, 2=450nm, 5=900nm. These data were extracted off the actual image labels.
Since we do not have acceptable darks (the periodic noise is the problem) I made a fake dark from the 999 msec dark frame (30860007) in a two step process. As before I looked at the row averages of the evens and odds separately. Did a linear fit to the data hoping to brute force the periodic noise into submission. From this I got column averages odds = 82.4 evens = 86.7. From these numbers I made fake dark current #1 - all even columns set to 82.4, all odd columns set to 86.7. Then I subtracted this from the real dark frame and looked at the residual - it wasn't optimal. So I empirically adjusted the odd even values til I got the mean of the real_dark - fake_dark down to -0.014 (std = 1.842). The values in this fake dark are odds=82.4, evens=87.4 (this is not very elegant I must admit). Of course this neglects the top to bottom change in the dark - however this is far below the periodic noise level. We definately need to do some thinking on this dark problem! Finally I ran all the images through the msi_lunar_calb sequence using this fake dark current. More to follow...
As you can see from the thumb nails above all the images have periodic noise to some degree or another. I am beginning an effort to quantify this... The plot below shows odd (bottom line) and even (top line) column averages, peak to trough is about 8 dn in this image. This noise makes getting a handle on the dark problematic. In the future we could possibly take a series (hopefully greater than 16) of identical darks (per given exp time) and then stack them in a cube - from this cube find the median (or possibly mode) per pixel and make a composite dark. Averaging does not work well unless you have a whole bunch of images, I think the median will help reduce the number of frames needed.
periodic noise plot 1>
FFT of image 30870000 (contains the comet).
fft of 30870000>
Jeff Warren, MSSS
Below is my attempt to correlate the image sequence table given on Jim Bell's (et. al.) comet observation planning page with the images described above. The image file names listed below are links to the FITS images found on APL's NEAR Science Data Center.
FITS # Exp Filt Compression Description ------------ -- --- ---- ----------- --------------------------------------- 30850007.FIT Unknown origin, partial image
30860007.FIT 01 999 7 Log+Fast 3.75x long dark at T = -30
30870000.FIT 02 999 0 None 1x Coma-centered clear image
30870010.FIT 03 250 0 Fast 2.5x Insurance against coma saturation 30870021.FIT 04 999 1 Fast 2.5x 550 nm coma image 30870032.FIT 05 999 2 Fast 2.5x 450 nm coma image 30870040.FIT 06 999 0 Fast 2.5x Coma-centered clear image 30870050.FIT 07 500 0 Fast 2.5x Insurance against coma saturation 30870065.FIT 08 999 5 Fast 2.5x 900 nm coma image
30870071.FIT 09 999 1 Log+Fast 3.75x 550 nm coma image 30870082.FIT 10 999 2 Log+Fast 3.75x 450 nm coma image 30870095.FIT 11 999 5 Log+Fast 3.75x 900 nm coma image
30870100.FIT 12 999 0 Fast 2.5x Coma NW mosaic position 30870110.FIT 13 999 0 Fast 2.5x Coma NW mosaic position 30870120.FIT 14 999 0 Fast 2.5x Coma NE mosaic position 30870130.FIT 15 999 0 Fast 2.5x Coma NE mosaic position 30870140.FIT 16 999 0 Fast 2.5x Coma SE mosaic position missing?? 17 999 0 Fast 2.5x Coma SE mosaic position 30870150.FIT 18 999 0 Fast 2.5x Coma SW mosaic position 30870160.FIT 19 999 0 Fast 2.5x Coma SW mosaic position
30870170.FIT 20 999 0 Log+Fast 3.75x Tail position 1; 2.5 deg. from coma 30870180.FIT 21 999 0 Log+Fast 3.75x Tail position 1; 2.5 deg. from coma 30870190.FIT 22 999 0 Log+Fast 3.75x Tail position 2; 5 deg. from coma 30870200.FIT 23 999 0 Log+Fast 3.75x Tail position 2; 5 deg. from coma 30870210.FIT 24 999 0 Log+Fast 3.75x Tail position 3; 7.5 deg. from coma 30870220.FIT 25 999 0 Log+Fast 3.75x Tail position 3; 7.5 deg. from coma 30870230.FIT 26 999 0 Log+Fast 3.75x Tail position 4; 10 deg. from coma 30870240.FIT 27 999 0 Log+Fast 3.75x Tail position 4; 10 deg. from coma
30870257.FIT 28 10 7 Log+Fast 3.75x short dark at T = -30
30870267.FIT 29 999(?) 7 Log+Fast 3.75x short dark at T = -30