UPDATE 9 DEC 2003 - CALIBRATED IMAGES (from which the mosaiced composite on the top of this page was made) NOW AVAILABLE: CLICK HERE.

SBIG ST-2000XM CCD Camera, graciously on loan from the Santa Barbara Instrument Corporation. The imaging program for the CCD camera was devised by J. Pasachoff and S. Sousa (Williams College) and was autonomously executed under computer control during the flight.  This program acquired green-filtered coronal images, executing a repetitive sequence of 10, 20, and 40 millisecond exposures with an inter-exposure cadence of 4 seconds throughout the total phase of the eclipse.
 

Notes on above Image:

1) The above image is a mosaic made from five (5) of the 40ms exposures taken as part of the CCD imaging program,

2) The camera platform pointing changed between images, thus providing a wider effective field-of-view.  The original raw images are shown in thumbnail form below:

3) The projection of the rotationally and translationally co-registered fields onto the sky is illustrated in two figures below.  The images are critically sub-pixel co-aligned on the features in the corona, not on the lunar disk. The apparent position of the moon moved significantly w.r.t. the Sun between exposures.
 

The figure on the left is a simple co-addition of the offset pointed raw images after co-registration.  As is apparent there is some amount of field overlap in portions of the wider mosaiced field.  The figure on the right illustrates the number of frames which contribute to a later done median collapse of the five images.  Black = 0, Dark Gray  = 1, Middle Gray = 2, Light Gray = 3.

4) Prior to combining the registered frames a calibration dark/bias frame was subtracted from each.  This calibration reference file was made from a median combination of 25 dark/bias frames taken on the aircraft immediately prior to the start of the eclipse imaging sequence.

5) NOTE: At this point the individual frames have *NOT* been flat-fielded, so the intra-frame photometric calibration is not correct, and they cannot acdtually be properly combined without introducing artifacts in the combination.  A flat-field reference file will very soon be provided by Steve Sousa and a proper image combination, devoid of the artifacts apparent in the topmost image can then be done.

6) Ahead of obtaining and applying a reference flat, the registered images were nonetheless median collapsed (though that will have to be done over again after flat fielding).  Because the CCD has a very large non-saturating dynamic range the information recorded cannot be shown in a single linear image.  Below, the inner regions of the corona captured in this image are shown:

To step through a linear display series, increasing the display gain by factors of 2 in each step, click on the above image. (This is a QuickTime movie; if your browser does not display it in-line then download and use a stand-alone QuickTime viewer).

7) More of the corona can be seen in a single image in a square root display.  Below are two flavors: Above inner and mid corona, and below mid and outer corona.
 

Again, the "seaming" artifacts will go away after flat-fielding.

7) More of the corona l structure can be seen in a single image in collapsing the dynamic display range by taking the Logarithm (base 10 in this case) of the image. Note, for example, the coronal streamer at the lower right which can be traced from the occulted solar limb (0.5 Rsun) out to 5 solar radii.

8) For better show the azimutally differentiated coronal features (streamers and holes) the underlying radial brightness gradient is partially subtracted.  This was done by measuring the radial brightness profile, on the azimuthal median, in all zonal regions one pixel wide, then subtracting 30% of the intensity of that radial profile at every corresponding radial zone.  30% is subjective, but the criteria is to not oversubtract any local feature to a negative value, which could then not be represented in a log display.  This "final" result is what was shown at the top of this page, and repeated here for contextual continuity:

The vertical striated "features" introduced in the corona on either side of the vertical axis of symmetry is due to display aliasing of the quantized subtracted radial profile.  This can be easily eliminated by simple 2 (or 3) point boxcar smoothing of the radial profile.  I certainly will do this after repeating this process with the data after it is flat-fielded.
 

Lastly, for now, on the CCD images, the same can be done for the 10ms and 20ms CCD images.  And then, when all are photometrically calibrated, all cam be combined onto a single final image to improve both the radial display range and the signal to noise.

As a FINAL note here, the actual data images are larger in linear scale by 4x, i.e., have 16x more pixels than shown above.  A smaller format was used just for web viewing.