Orientation and Scale Verification and Correction

A standard array orientation was adopted for all images with axes orthogonal (within a few degrees) with celestial north and east, and displayed in such a way to place celestial north to the top of the image, and east to the left.  The relevant PDS3 keywords and adopted values are:

  LINE_DISPLAY_ORIENTATION=UP
  SAMPLE_DISPLAY_ORIENTATION=RIGHT
  CELESTIAL_NORTH_CLOCK_ANGLE=0

The alignment was achieved with label revisions and simple array transformations: axis mirroring, 90, 180, or 270 degree rotations.  These transformations affect the data order, but preserve the data values.  Images with axes that are not orthogonal to north and east may have also been revised.  All images were given the same line and sample display orientations, mirroring the array and updating the CELESTIAL_NORTH_CLOCK_ANGLE, as needed.  The keywords SENSE and CROTA1 in the FITS formatted headers, indicating the position angle of the sample axis, were also revised to match the new orientations.

Three methods were used to verify and correct the image orientations:

  (1) Background star patterns: For all images, the background stars were examined and compared to a Digital Sky Survey image centered at the ephemeris position of the comet.  This method provides the most robust estimate of the orientation and pixel scale.  However, the background stars are not always present in sufficient numbers for success.  In addition, background artifacts, image coadding, or multiple sky exposures on the same plates can cause confusion.

  (2) Comet proper motion: The proper motion of the comet may be estimated over a sequence of images and compared to the comet's ephemeris from JPL HORIZONS.  For single images, aligning the image with the comet's proper motion contains an uncertainty of 180 degrees, so it must be combined with an analysis of the tail orientation.

  (3) Comet tail direction and coma shape: Comet tails tend to lie near the projected Sun-comet vector, and can be used to break the 180 degree degeneracy in the proper motion alignment for single images.  However, being a near-nucleus data set, the tail direction was not always obvious.  In such cases the sunward direction, which tends to have a parabolic shape, can also be used to infer orientation.

Some images could not be verified on their own (e.g., independent narrow-band images with little continuum flux and few to no stars).  In such cases the orientation has been assumed to be the same as any contemporaneously obtained continuum or broad-band images from the same observatory.

Of the 3523 images, 857 required an x-flip, 2976 images were rotated by 90, 180 or 270 degrees, and 6 images could not be oriented (listed in know-issues.asc) and are marked with CELESTIAL_NORTH_CLOCK_ANGLE=UNK.

Timing and pixel scale errors were occasionally discovered while measuring image orientations with the stars.  Timing errors were revealed when the comet was not found at the ephemeris position for the stated time.  Small errors, on the order of minutes to tens of minutes, were generally not obvious, except when the data were placed in time sequence and the motion of the comet showed the images were out of order.  The observation times of eight images were revised, summarized in the document time-revisions.asc and listed in revision-summary.asc.  26 images had times that were duplicated in another image, commonly due to low time-keeping precision.  These are documented in known-issues.asc and the NOTE field of the labels.

Pixel scales were derived from FITS header keywords PLTSCALE and APSIZE, or the FITS comments.  The latter was important for identifying rectangular pixels.  When aligning with stars, some pixel scales were clearly incorrect.  These have been updated with improved estimates, derived by hand.  All 248 pixel scale updates are presented in the revision-summary.asc document.