On the the stability of images grabbed with a Matrox Meteor II frame grabber card
Detlef Koschny, 28 Sep 1998
Introduction At ESA/SSD, we currently work on the analysis of meteor observations obtained with an image-intensified video camera. One of the goals we have is to perform astrometry of a complete meteor event, i.e. from the image of the meteor on several individual frames (Note that the video camera obtains 50 half frames per second. Depending on the grabber and the software, these may be combined conforming to the definition of PAL video to 25 full frames per second). Currently, we use the software Astrometrica to perform astrometry from sum images (a sum image is obtained by taking the maximum value of each pixel for all frames of the event and storing it in one frame). When measuring individual frames, one of the questions that arises is whether it increases the accuracy of the measurement if the reference stars are measured in each individual frame, or if it is sufficient to assume that their position did not change from one frame to the next. Since we use wide angle optics and short exposure times (<1/50 s), the movement of the sky due to the rotation of the Earth is negligible. Atmospheric turbulence (szintillation) results in a fluctuating stellar image, i.e. both the diameter and the maximum brightness of the image changes a little bit. However, the center of the disk should not change, the typical image shifts introduced by turbulence are much less than our pixel resolution (40"). However, both the stability of the image on the analog video tape and the reproducability of the grabbing process of the frame grabber card are not known. I therefore looked at the stability of the position of a star image in several grabbed sequences. Procedure Three different image sequences (called event) were grabbed with a Matrox Meteor II frame grabber card used in a 160 MHz Pentium PC with 32 MB RAM. The video recorder used was a Panasonic NV-HS900. The events were grabbed from a video tape obtained by the camera VideoMet at an observing campaign during the Geminids on 14 Dec 1996. One event had 40 consecutive frames, the other two had 24 consecutive frames. We read the individual frames into an IDL program which displays a small window of the total frame with large magnification to get an impression of the star image. The window was chosen manually to be centered around a star of about 4.5 mag. Figure 1 shows an extremely magnified image of the star for the four different cases that were analysed. Note that for the display the IDL routine rebin was used which smooths the image. The pixel coordinates and time on the tape for the different events are shown in Table 1. In the third sequence, we analyzed two different stars.
Figure 1: Enlarged star images, shown with the IDL function rebin. See Table 1 for a description.
Table 1: Parameters of the four analyzed cases.
Image in Figure 1 A B C D Upper left Upper right Lower left Lower right
Time of scene in tape counter in hh:mm:ss 00:51:36 00:54:41 00:30:14 00:30:14
x/y of lower left corner in pixel 275/320 250/330 150/190 380/240
x/y of upper right corner in pixel 305/350 280/360 190/220 420/280
For the cases mentioned above, we used the IDL routine cntrd which can be downloaded from the GSFC IDL astro library at http://idlastro.gsfc.nasa.gov. This routine finds the center of a star image. For each individual frame within one event, the difference in the center position with respect to the first frame was determined.
Results and interpretation The following figures show the movement of the center position of a star with respect to the first frame for the different cases A to D as defined in Table 1. Cases A and B both show a movement of less than one pixel. The shift in y-direction is about a factor of two smaller than the shift in x-direction. It is not clear whether this has to do with the fact that our images were composed of two half frames be reminded that in the video format, one half frame corresponds to the even pixel rows, the next half frame to the odd pixel rows. If the center of the star shifts on a time scale of 1/50 s, we might actually average out the shift in x-direction by putting two half frames together. This could be tested by looking only at half frames, however, this was not done yet. Cases C and D are a little bit different there is a shift of several pixels in x-direction. Both cases are stars from the same event at a time marker of 00:30:14. Both cases show the shift in a similar but not identical way. The event was grabbed directly after the VCR was switched on. On the screen, overlaying the star image, is a display from the recorder. Quite possibly this causes the jump in the image. The fact that both stars show this jump means that the complete image is moving. However, they are not exactly identical. Thus, there must be an additional rotation involved. To verify that this jump was caused by the startup of the VCR, we grabbed the scene again.

Figure 2: Relative position of star center in pixels, case A.
Figure 3: Relative position of star center, case B.
Figure 4: Relative position of star center in pixels, case C.
Figure 5: Relative position of star center in pixels, case D. Note that the first center position is at 0/-1.
Figure 6: Relative position of star center, case E: same event as case C, but grabbed again.
Conclusions The shift of the center of a star during one event is less than one pixel. We therefore conclude that it is acceptable to add several consecutive images without correcting for any shift. It should also be acceptable to perform astrometry on several consecutive frames with the reference stars determined only in the first frame.
Appendix Here is a listing of the IDL code used to look at the data.
;file starjump ; displays one star image to determine whether stars jump - based on ; a read routine for ViDAS bmp files written by Joe Zender ;initializing variables ;The files are stored in seq1_0000, seq1_0001, etc. The numbers are ;counted up in the for loop lateron. For a different event, seq1 is ;replaced by seq2 etc. fileName="z:\siv\bmpseq\seq1_" x0 x1 y0 y1 = = = = ;find a star manually, put a window around it

zoom = 10 fwhm = 10

;zoom factor for the display on the screen ;full width at half max in pixel in original image
dx = x1 - x0 dy = y1 - y0 x_cen = FLTARR (40) y_cen = FLTARR (40) window,0,xsize=dx*zoom,ysize=dy*zoom inp='x' for i=0,MaxNoOfFrames do begin name = fileName + string(i, format='(I3.3)') + ".bmp" print, "opening file " + name res = read_bmp (name) res = res [x0:x1-1, y0:y1-1] ;-1 to get even no. of pixels image = rebin (res, x*zoom, y*zoom) ;enlarge the image tvscl, image cntrd, res, x/2, y/2, x_help, y_help, fwhm image) ;find centroid (in original
;cntrd is from the gsfc idl lib. x_cen (i) = x_help y_cen (i) = y_help ;can't put these directly in the above line end ;determine delta to first position x_cen_0 = x_cen (0) y_cen_0 = y_cen (0) ;note: MaxNoOfFrames was set interactively to the current no. of frames for i=0,MaxNoOfFrames do begin x_cen (i) = x_cen (i) - x_cen_0 y_cen (i) = y_cen (i) - y_cen_0 end ;generate another window to plot the x/y position relative to first pos. window, 1, xsize=500, ysize=500 plot, x_cen, y_cen, psym=-2, xrange=[-2,2], yrange=[-2,2], ticklen=1 END

AMPHITHEATRE - TECHNICAL DETAILS AMIRAUTE HOTEL DEAUVILLE
Sound Equipment : Public Address system :JBL loudspeakers, Bose 100v FreeSpace 102F Loudspeaker, QSC 100v Line amplifier, Power amplifier C.Audio, Sound is routed through a BSS Soundweb DSP system controller. JBL monitor speakers, on stake and mobile. 4 JBL left and right speakers type SIM 200 watt. 8 Subwoofers JBL type S1S 400 watt, 2 JBL central speakers type S1C 200 W, JBL surround sound with THX AC3 DTS processor. 4 HF hand held SENNHEISER SKM 5000 microphones, frequency 750,600 773,500. K&M Telescopic mike stand with boom. 32 channel monitor mixer ALLEN et HEATH GL 3300. Audio and Video equipment: Tascam double cassette deck 202 MK3, Tascam High Resolution DA-45HR DAT recorder, PANASONIC A 360 DVD player, Yamaha SPX-1000 effects processor, LCD Proxima Pro AV 9310 Projector for video and computer display ( LCD screen 3" TFT active true XGA 1024x768 ; SXGA (1280 x 1024) via compression, output 2100 LUMENS ANSI), long focus zoom for DP 9310 PROXIMA, Extron 8 x 4 RS232 Matrix switcher A/RGBHV Extron 12 x 8 RS232 Matrix switcher A/V. Universal computer interface ANALOG WAY Splitter , Sony SP Betacam video recorder/playback, PANASONIC NV HS 900 VHS/S-VHS Video recorder, 1 x 10 VITY Stereo audio/video distributor. Screen: Oray electric 6 m x4.5 m projection screen CD player: Sony 300 disc player Lighting : 15 RVE 1kw shuttered spots, 10 RVE 1 kw PC floods 2 2kw RVE PC floods, 25 PAR 64 lamps. Stage: 112 m stage , rear and front of stage curtains Lighting console: 96 3Kw circuits by Light Technology Freight lift: At back of stage Length: 5.56m, Width: 2.40 m, Height : 2.10 m (L216in x W 94in xH82in) Maximum load: 3 tonnes (6600 lbs) At the centre back of the stage, (opening in the width direction) Access to the lift via a locked garage (parking space for 2 vehicles) Several vehicles or lorries can gain access in front of the garage Fixed hoist total maximum weight 500 kg (1100 pounds) uniformly distributed 112 square metre stage (9,9m x 11,4m - 32ft8in X 37ft7in) with a 6m x 4.5 m screen on mobile base. All this equipment is included in the hire of the theatre, and will be exclusively operated by our technician from the fixed main control room. For all other equipment, a pre-wired extension control centre located at balcony level is available. Supplementary equipment can be hired from us subject to quotation. For more precise details either on the equipement or the technical characteristics, you can contact Damien (Theatre technician) on +33 (0)2.31.81.82.83.