8.4: Technical Requirements for DEIMOS Image Display
A typical DEIMOS frame which includes some prescan and overscan
information from all 8 CCDs will consist of about 8500 columns and
8200 rows. As 16-bit integers this requires nearly 140 Megabytes of
storage, and when converted to IEEE floating point this storage
doubles.
8.4.1: Requirements for the Data Acquisition System
The demands of the image display exceed those of many existing
systems. For the sake of speed and for other considerations this
places certain extra requirements on the CCD data acquisition system
(DAS) (described in Chapter 6). These requirements are reiterated here
beccause they are driven by the needs of the image display.
8.4.1.1: Mosaicking Needs
For typical imaging cases the pixels will be read from 4 CCDs,
and in spectral mode pixels will be read from 8 CCDs.
For quick look and for the sake of pre-existing image display tools
these image sections will be mosaicked as they are stored in memory.
The software which receives the streams of pixel values from the CCDs
should be able to do the following:
-
in advance of the readout determine the overall size and
layout of the various sections which comprise the image
-
create FITS header keywords which document the layout of
these image sections
-
be cognizant of the original CCD, amplifier, row, and column
of each pixel as it arrives
-
calculate offsets and strides through memory as needed to
put each pixel in its place in memory
-
create a FITS header for the overall mosaic in memory
-
create a FITS header for the image section from each CCD,
and maintain stride information necessary to access those pixels
-
create a FITS header for combined image sections from
adjacent CCDs with logically related data,
and maintain stride information necessary to access those pixels
When the images are written to FITS files on disk the user should have
the option of storing them as:
-
a single FITS file containing the entire mosaicked image with
FITS header cards that document the image subsections
-
one FITS file for each CCD with FITS Grouping Convention
cards that point to the FITS files from the other CCDs
-
a FITS file for each logically related set of pixel data
with FITS Grouping Convention
cards that point to the FITS files from the other CCDs
8.4.1.2: Slitmask Alignment Exposures
The slitmasks require that the telescope be precisely pointed
such that the light from each object passes through its slit.
Each slitmask will have a number of rectangular apertures through which
it will be possible to observe alignment stars (see Chapter 3).
In order to avoid
sacrificing large regions of slitmask the alignment stars will typically
be chosen as close together as possible and clustered along lines
parallel to the dispersion direction.
-
the alignment apertures should be rectangular and closely
aligned with the CCD axes
-
the alignment apertures should have the same size
-
the alignment apertures should project to approximately 50
pixels on a side (4 arcsec)
-
the alignment apertures should not overlap
-
the CCD readout regions surrounding each aperture may overlap
and require pixel duplication in the resulting image
-
there may be up to 16 alignment apertures on a slitmask
These constraints permit the alignment images to be handled by the
same software used for normal CCD readouts.
Only a small fraction of the CCDs will be usefully illuminated so
the DAS need not readout the entire imaging area.
The slitmask alignment exposures do not require calibration by prescan
or overscan pixels; they should not be stored or displayed.
In order to accelerate the readout the DAS must be able to do the
following:
-
obtain information from the slitmask database which describes
the location of the alignment apertures
-
rapidly readout only those rectangular sections of the CCDs which
surround the alignment apertures
-
place the sections of image data into a mosaic sorted by
X-position along the slitmask
-
document these image sections with FITS keywords
(in much the same manner as for
sections from multiple amplifiers on multiple CCDs)
8.4.1.3: Gathering of Image Statistics
Several astronomers have expressed interest in seeing some kind of
quick statistical analysis of the image data as it is first displayed.
This interest was prompted the fact that the displayed images may be
binned by a factor of 4 to 8. With such binning it will be difficult
for astronomers to perform a manual scan of the images for defects.
As the sections of the image data arrive from the CCD controllers
the data acquisition system should accumulate some simple
statistics. Some of these will assist the astronomers to make
data quality assessments; others will serve to accelerate the
operation of the image display software.
Subject to the constraint that the calculations do not significantly
delay the readout and storage of the CCD image, the following
statistics should be measured for each separate CCD amplifier:
-
Number of pixels with DN value above a user-settable
``saturation'' level
-
Number of pixels with DN value below a user-settable
``base'' level
-
The maximum and minimum DN values within the image data section
-
The maximum and minimum DN values within the calibration
(pre- and over-scan) data sections
-
The mean DN value, RMS deviation, and mode within the
calibration data sections
-
The mode of the DN values within the image data section
-
The number of columns whose mean DN value is below the
mean level of the calibration data
-
The number of rows whose mean DN value is below the
mean level of the calibration data
-
The mode of the DN values within the image data section
8.4.2: Requirements for the Image Display Graphical User Interface
In previous Keck detectors (i.e., HIRES and
LRIS) the detectors have been single CCDs read through multiple
amplifiers. The pre/overscan data have been located outside the
bounds of the image data. This permitted the image display algorithms
to consider only the central block of pixels when reducing the image
from 16-bit integers to an 8-bit (or less) grey-scale. In DEIMOS not
only will there be overscan data in the midst of the image, but in
both spectral and imaging frames there will be regions of CCDs which
are not illuminated.
8.4.2.1: Overview of Hardware and Protocol requirements
Single monitors that can display 8k x 8k images do not exist.
Monitors that can display 2k x 2k pixels do exist, but they are quite
expensive and will typically not be available at remote sites.
``Monitor walls'' can be built by stacking arbitrary numbers of commonplace
monitors, but again these will not be available at remote sites.
We choose to work within the bounds of typical existing displays which
have about 1k x 1k pixels and 8-bit color depth. If circumstances
provide a larger or deeper display then the display should be able to
take advantage of them.
-
The image display tool should use the X Window System, Version 11
as the underlying protocol for image output and user input.
-
If the X server and the image display clients are on the
same system then the X11 protocol should make use of the MIT
shared memory extensions to accelerate the image display.
-
Typical image manipulations must be possible on screens with a
resolution of about 1k x 1k pixels.
-
The pan, zoom, and main image display widgets must be
separable into distinct windows. Each of these sections of
the display must permit its size and location to be
independently configured. The image display client must be
able to place these different windows onto different X11
DISPLAYs.
-
The image display tool must be able to recognize mosaicked
image section information in the FITS headers which describe
up to 16 disjoint rectangular sections of CCD image
information. Examples: one section from each amplifier
during readout of all CCDs; or, multiple disjoint sub-rasters
during readout of mask alignment exposures.
-
The image display tool must recognize the simple statistical
information placed in the FITS headers by the DAS. These
keywords will contain the information typically needed by a
display client (e.g., image limits, bias levels, histograms)
to choose its color look up table. If those keywords exist
the GUI should display the image using their values and avoid
re-doing those calculations over the entire image.
-
The image display tool should provide a GUI by which the user
can obtain values of the image statistics inserted by the DAS;
e.g., to see the number and locations of of saturated pixels,
etc.
-
The image display tool must be able to use the image section
information to selectively ignore regions of the image which are
composed of prescan or overscan pixel data.
-
The image scaling algorithms should be capable of handling a
multi-modal distribution at the dark end which consists of overscan,
un-illuminated, and sky pixels.
-
The GUI should be able to recognize World Coordinate System
(WCS) information in the FITS headers which describe up to 16
disjoint rectangular sections of CCD image information.
-
The GUI should permit blinking between multiple images.
There should be mechanisms which permit the user to
choose a separate color LUT for each of the images.
When one or more images are zoomed on a sub-raster there
should be a mechanism which permits the user to specify
whether or not the other blink images use the same
zoom.
-
It should be possible to obtain an annotated printout of
any data displayed in the GUI.
8.4.2.2: Display of World Coordinate System Information
The GUI will utilize WCS information in the FITS
headers. As the mouse tracks over the image it will provide a real-time
update of the coordinate position on the sky (for image frames) or the
slitlet/object and observed wavelength information (for spectral frames).
The precise WCS mappings for DEIMOS will be highly nonlinear
and may exceed the capabilities of the simple cases described in the
FITS WCS draft.
The real-time tracking must make use of the simple components of the
WCS mapping as described in the main body of the WCS draft.
If the fully accurate and nonlinear WCS mapping can be computed fast enough
it may supercede the simple transformation; otherwise upon
user request the GUI should perform the mapping to full accuracy.
For any kind of frame the user should have the option of displaying
WCS information in any one of the following coordinate systems:
-
X and Y within the entire mosaicked FITS image [pixel]
-
The identity of the particular CCD plus X and Y within that CCD [pixel]
-
In the case of multi-amplifier readout of any CCD, the
identity of the particular amplifier of a particular CCD plus
X and Y within that CCD [pixel]
When the cursor is positioned over pixels that contain calibration
(prescan or overscan) data the position display should revert to
indicate the X and Y pixel data. It will simultaneously
indicate the type of the calibration pixels.
For imaging frames the user should have the option of displaying WCS
information in any of the following coordinate systems:
-
Geocentric Apparent Right Ascension and Declination
-
Catalog Right Ascension and Declination
-
Catalog Right Ascension and Declination with an added offset
specified by the observer
-
Deviation in catalog Right Ascension and Declination from a
reference point specified in the slitmask catalog database
-
Catalog Galactic Longitude and Latitude (System II) [degree]
-
Slit Mask coordinates X and Y [mm]
For spectral frames the coordinate system will be known to high
precision because of the requirements for aligning the slit masks.
The user should have the option of displaying WCS
information that tells the following:
-
Slitlet ID and observatory wavelength
-
Name and catalog Right Ascension and Declination of the object(s) in
the slitlet
-
Catalog Right Ascension and Declination of any position along
the slitlet and observatory wavelength
-
Slit Mask coordinates X [mm]
Steve Allen <sla@ucolick.org>
$Date: 1996/03/19 06:29:41 $