Data Set Overview
    =================
      The Mercury Dual Imaging System (MDIS) consists of two cameras, a
      Wide Angle Camera (WAC) and a Narrow Angle Camera (NAC), mounted on
      a common pivot platform. This dataset includes the Calibrated Data
      Record (CDR) version of all valid, calibratable images acquired
      during the cruise phase to Mercury and the Mercury orbital phase,
      and includes post-launch checkout images, flyby images of Earth,
      the Moon, Venus, and Mercury, in-flight calibration images, and
      images taken during Mercury orbit including those targeted at
      Mercury, or taken during imaging of comets or searches for Mercury
      satellites or vulcanoid asteroids. This CDR dataset is the product
      of conversion of raw data (Experiment Data Records, or EDRs) to
      one of two options:
 
         physical units of radiance, or
         radiance factor or I/F
 
      as indicated by the filename:
 
      'pcrnnnnnnnnnf_tt_v', where
 
         p = product type = C calibrated
         c = camera (W WAC or N NAC)
         r = spacecraft-clock-partition-number minus 1 [0, 1]
             for pre- or post-spacecraft-clock-reset
         nnnnnnnnn = Mission Elapsed Time (MET) counter taken from the
                      image header (and same as original compressed
                      filename from SSR). NOTE: this is a spacecraft clock
                      seconds counter, and the value in the filename
                      corresponds to the LAST second of the exposure.
         f = Filter wheel position (A, B, C, D, E, F, G, H, I, J, K, L, U)
             for the WAC.  It is M for the NAC, which has no filter wheel.
             It is U if the position is unknown.
         tt = data type (RA radiance, IF for NAC I/F or WAC I/F corrected
             for time variations in responsivity, or IU for WAC I/F
             uncorrected for time variations in responsivity)
         v = version number
 
      The data exist in a format paralleling that of the raw data: images
      have the same line and sample dimensions, and information in
      attached PDS labels is updated to convey differences in units and
      processing history. This dataset also includes ancillary data files
      that tabulate the contents of the volume and documentation files.
 
      For more information on the contents and organization of the volume
      set refer to the aareadme.txt file located in the root directory of
      the data volumes.
 
 
    Versions
    ========
 
      Version numbers of CDRs increment when there is a change to the
      content of the calibrated image data, typically resulting from an
      update to radiometric calibration.
 
      Version 1 was the prototype version of the CDRs produced internally
      to the MDIS team for validation.
 
      Version 2 was the first publicly released version of MDIS CDRs, and
      was first released in MDIS PDS release 3 which contains data through
      Mercury flyby 1. In version 2 the labels were modified to include
      additional information on image statistics, and to replace the values
      in un-calibratable pixels (saturated pixels, pixels under the dark
      strip at the edge of the CCD) with special values. In addition the
      calibrations of WAC filters 3 and 6 (C and F, at 480 and 433 nm)
      were adjusted from the version 1. See the discussion
      below for description of the calibrations applied to CDR images,
      or the file CALINFO.TXT in the CALIB directory for version histories
      of each of the calibration matrices. Version 2 CDRs were produced
      using:
        - version 0 of the dark model;
        - version 4 of the flat-field correcton (except version 2 of the
          flat field correction in WAC filter 2 and NAC binned);
        - version 2 of the WAC responsivity correction (except version 3
          in filters 3 and 6), and version 1 of the NAC responsivity
          correction; and
        - in I/F images, version 0 of the solar spectrum.
 
      Version 3 CDRs were first released in MDIS PDS release 4 containing
      data through Mercury flyby 2. In version 3 there were no changes to
      labeling of CDRs, but the responsivity correction was updated. The
      previous linear model of the effect of CCD temperature was replaced
      with a quadratic function. In addition the calibration of the NAC
      was adjusted. Version 3 CDRs were produced using:
        - version 0 of the dark model;
        - version 4 of the flat-field correction (except version 2 of the
          flat field correction in WAC filter 2 and NAC binned);
        - version 4 of the WAC responsivity correction, and version 3 of
          the NAC responsivity correction; and
        - in I/F images, version 0 of the solar spectrum.
 
      Beginning in MDIS PDS release 5, which contains data through Mercury
      flyby 3, labeling of CDR images was updated in response to an MDIS
      flight software upload in August 2009. Several additional items of
      housekeeping were added to the headers of downlinked images, and
      propagated into the CDR labels. No changes in radiometric
      calibration were made at this time
 
      Beginning in MDIS PDS release 6, which contains data through the
      beginning of Mercury orbit, three additional items were added to
      CDR labels providing information on timing and image targeting. One
      small change was made to radiometric calibration to null a problematic
      column of image data.
 
      No changes of these types pertain to PDS releases 7 or 8 which contain
      data from months 3-12 of Mercury orbit.
 
      Beginning in PDS release 9, which contains data from months 13-18 of
      Mercury orbit, version 4 CDRs are first released. In this update,
      five further corrections and additions to radiometric calibration are
      implemented. (a) A small error in calculation of the correction for
      frame transfer smear is corrected. (b) The flat-field correction for
      WAC filters 3 and 6 is updated, using normalized median values of
      several hundred images acquired during orbit, to which a photometric
      correction is applied to remove a gradient to illumination across the
      images. This retires a problem with the flat-field originating
      from the ground calibration. (c) The WAC temperature-dependent
      responsivity is updated using median values of images covering a
      narrow range of photometric angles, in the period prior to the
      change in WAC performance described in the following discussion.
      (d) A new term is added to the calibration equation, 'Correct',
      to correct for time variation in responsivity in all WAC filters.
      Beginning 24 May 2011, later analysis indicated that transmission
      decreased by different amounts in each filter, most in the shortest-
      and longest-wavelength filters. Subsequently response increased
      slowly over several months approaching previous values. This is
      believed to result from deposition and subsequent bakeoff of a
      contaminant on the MDIS outer optic. The deposition may have
      resulted from MESSENGER's first period with periapsis of its orbit
      near the sub-solar longitude at planet perihelion. The correction
      is maintained as a look-up table, derived empirically by comparing
      median properties of WAC images taken within a narrow range of
      photometric geometries. (e) Following the apparent recovery of
      responsivity about 10 months after the 'contamination event',
      response in several WAC filters - especially 6 and 7 (F and G) -
      began to fall again relative that response through other filters.
      It was determined that the rate of falloff in response was related
      to the amount of time the filter was 'in position' for imaging;
      generally MDIS is pointed in the direction of the illuminated
      surface of Mercury. It is speculated that intense radiance from
      Mercury caused loss of transmission in the affected filters. A
      correction is maintained as part of the same look-up table derived
      to track dissipation of the contaminant.
 
      Version 4 CDRs are produced using:
        - version 0 of the dark model;
        - version 5 of the flat-field correction in WAC filters 3 and 6;
          version 2 of the flat field correction in WAC filter 2
          and NAC binned, and version 4 in all other WAC filters whether
          binned or unbinned;
        - version 3 of the NAC responsivity correction; version 5
          of the WAC responsivity correction; and
        - version 3 of the temporal correction to WAC responsivity
            'Correct'; and
        - in I/F images, version 0 of the solar spectrum.
 
      In PDS release 15, the final end-of-mission data release which
      includes all data, version 5 CDRs are released. In this update,
      three further corrections and additions to radiometric calibration
      are implemented. (a) The flat-field correction for several WAC
      filters in non-binned and binned states is updated, using the same
      approach of normalized median values of low-contrast field filling
      images acquired during orbit, except using more images less
      affected by compression than in version 4 CDRs. (b) The NAC and
      WAC temperature-dependent responsivity are updated based on
      further analysis of images acquired at differing temperatures.
      (c) The 'Correct' term in the calibration equation is updated
      based on extensive analysis of overlaps of images acquired at
      different times, and covers the entire period of Mercury orbit.
 
      Version 5 CDRs are produced using:
        - version 0 of the dark model;
        - version 6 of the flat-field correction in WAC filters 3 and 6;
          version 2 in WAC filter 2 and NAC binned; version 6 in binned
          images in WAC filters 3, 4, 5, 6, 7, 9, and 12; and version 4
          in all other WAC filters whether binned or unbinned;
        - version 4 of the NAC responsivity correction; version 6
          of the WAC responsivity correction;
        - version 5 of the temporal correction to WAC responsivity
            'Correct'; and
        - in I/F images, version 0 of the solar spectrum.
 
 
    Parameters
    ==========
      MDIS observing scenarios are constructed using a set of key
      variables ('configurations') which include the following. All, with
      the exception of filter selection, are available for both the NAC
      and the WAC. Only the WAC has selectable filters. The imagers can
      only be used one at a time.
 
      Compression: MDIS images may optionally be compressed in a number
      of ways, including pixel binning on-chip, 12-8 bit compression, and
      FAST/Differential pulse code modulation (DCPM) lossless
      compression, all carried out in the instrument hardware; further
      pixel binning, subframing, wavelet compression, and jailbars, are
      all carried out in the spacecraft Main Processor (MP), either
      individually or in combinations.
 
      Pixel Binning: MDIS images can undergo 2x2 pixel binning in the
      focal plane hardware (also known as 'on-chip' binning), resulting
      in a 512 x 512 image. Images can also be compressed using the MP,
      either in addition to DPU binning, or instead of DPU binning. MP
      pixel binning options of 2x2, 4x4 or 8x8 are available.
 
      12-8 bit compression: Images are read off the detector in 12-bit
      format. 12 bit images may converted to 8 bit images using lookup
      tables (LUTs) designed to preferentially retain information at low,
      medium, or high 12-bit DN values. Eight LUTs are available, and
      shared between the NAC and WAC.
 
      FAST/DPCM compression: All images are compressed losslessly using
      FAST/DCPM compression as they are read out of the DPU, to conserve
      recorder space. Once the data are written to the recorder, they can
      be uncompressed and recompressed more aggressively in the MP.
 
      Wavelet compression: Images may be integer wavelet transform-
      compressed in the MP, typically at 3:1 for color data and 4:1 for
      monochrome data, but any value from 1 to 32 can be used.
 
      Subframing: In order to manage downlink resources, up to 5 portions
      of the image can be selected in the MP and downlinked as subframes.
      The subframes are allowed to overlap. During EDR construction the
      subframes are all mosaicked into the original image, and this
      reconstruction is carried through to the CDR level.
      During orbital operations this option is not regularly used.
 
      Jailbars: Intended for data management during optical navigation,
      jailbars are selected columns of an image retained by the MP.
      Commanded column spacing values are not restricted and can be
      set to any integer value between 1 and 1024, but the spacing is
      fixed throughout the image. During EDR construction the
      jailbars are all mosaicked into the original image, and this
      reconstruction is carried through to the CDR level.
      During flight this option is not regularly used.
 
      Exposure Control: The exposure time of MDIS images can be set
      manually by command, or automatically by the software. In manual
      mode, exposure times from 1-989, 1001-1989, ..., to 9001-9989 ms
      are available. In autoexposure mode the exposure time of the next
      image is computed by the DPU software, and cannot exceed 989 ms
      in duration. If the time of the next image occurs before the
      calculation can be completed, and pixel binning or filter
      position change, then the algorithm compensates for predicted
      changes in scene brightness and filter transmission using an
      onboard data structure.
 
      Pointing: The MDIS imagers are mounted on a pivot platform, which
      is itself mounted to the MESSENGER spacecraft deck. The pivot
      platform is controlled by a stepper motor, which is controlled by
      the Data Processing Unit (DPU). The pivot platform can move in
      either direction. The total range of motion is 240 degrees, limited
      by mechanical 'hard' stops, and is further constrained by 'soft'
      stops applied by the software. The nominal pointing position for
      MDIS is defined as 0 degrees, aligned with the spacecraft +Z axis
      and the boresight for several other instruments. The range of the
      soft stops is set to 40 degrees in the spacecraft -Y direction
      (toward the MESSENGER sunshade) and +50 degrees in the +Y direction
      (away from the sunshade). The pivot position can be commanded in
      intervals of 0.01 degrees within this range.
 
      Filter selection: The WAC imager contains a 12 position filter
      wheel to provide spectral imaging over the spectral range of the
      CCD detector. Eleven spectral filters span the range from 395 to
      1040 nm, while the twelfth position is a broadband filter for
      optical navigation.
 
 
    Processing
    ==========
      An MDIS image downlinked by the spacecraft unpacks into a
      succession of one or more compressed image subframes with binary
      headers containing housekeeping items that contain full status of
      the instrument hardware, including imager, software configuration,
      temperature, voltage, and current readings, pivot position, and a
      time stamp.
 
      The data in one CDR consists of a single reconstructed image frame,
      with the accompanying header translated into text format in the
      label, and converted in physical units. Each frame has dimensions
      of spatial samples in the form of detector columns and detector
      rows. Each frame is formatted into one file (suffix *.IMG),
      with an attached label.
 
      Each image has dimensions XX pixels in the sample dimension and YY
      pixels in the line dimension, where:
 
      XX (columns) = 1024/binning, where 1024 is the number of columns
      read off the detector, and binning is 2, 4, or 8 (the product of
      binning at the instrument level and by the MP).
 
      YY (rows) = 1024/binning, where 1024 is the number of columns read
      off the detector, and binning is consistent with the sample
      binning.
 
      Subframes are not retained as separate entities but are reassembled
      into their original coordinates in the image. Parts of the
      reconstructed image not included in the subframes are given a value
      of zero.
 
      The label contains all of the housekeeping information in their raw
      form. Selected ones are listed in duplicate in calbrated form (e.g.
      degrees Celsius, volts, or amperes. In addition, value-added
      information in the label describes image pointing and objectives.
 
      The relationship of radiance in CDRs to raw data in EDRs is
      described by the following equation:
 
      L(x,y,f,T,t,b) = Lin[DN(x,y,f,T,t,b) - Dk(x,y,T,t,b) -
      Sm(x,y,t,b)] / {Flat(x,y,f,b) * t * Resp(f,b,T)}
 
      where:
 
      L(x,y,f,T,t,b) is calibrated radiance in units of
      W / (m**-2 microns**-1 sr**-1), measured by the pixel in column x,
      row y, through filter f, at CCD temperature T and exposure time t,
      for binning mode b. F, T, t, and b are all given in the label.
 
      DN(x,y,f,T,t,b,MET) is the raw DN measured by the pixel in
      column x, row y, through filter f, at CCD temperature T and exposure
      time t, for binning mode b, and Mission Elapsed Time (MET).
      f, T, t, b, and MET are all given in the label.
 
      Dk(x,y,T,t,b,MET) is the dark level in a given pixel, derived from
      a model based on exposure time and CCD temperature,
 
      Sm(x,y,t,b) is the scene-dependent frame transfer smear for the
      pixel,
 
      Lin is a function that corrects small nonlinearity of detector
      response,
 
      Flat(x,y,f,b) is the non-uniformity or 'flat-field' correction,
 
      Resp(f,b,T) is the responsivity, relating dark-, flat-, and
      smear-corrected DN per unit exposure time to radiance,
 
      and
 
      t is the exposure time.
 
      The above equation assumes that data are in the native 12-bit
      format in which they were read off the CCD, and that onboard
      application of 12-to-8 bit lookup tables (LUTs) has been inverted.
 
      This correction is done step-wise using the calibration tables and
      images in this directory as follows.
 
 
      (1) Inversion of 12 to 8 bit Compression
      ========================================
 
      8-to-12 bit inversion of DN values is required when the data are
      8-bit. There are 8 inverse lookup tables (LUTs). The table to use
      is indicated by the choice of forward LUT that was applied, as
      indicated by the keyword MESS:COMP_ALG whose value is
      from 0 through 7. An 8-bit value (in a row of the table) is
      inverted by replacing it with the 12-bit value in the column
      corresponding to a particular LUT.
 
      (2) Subtraction of modeled dark level
      =====================================
 
      There are four separate models of dark level (dark current plus
      electronics bias), for the MDIS-WAC and MDIS-NAC (as indicated
      by the keyword INSTRUMENT_ID), and for each camera, without pixel
      binning turned on (MESS:FPU_BIN = 0) or with pixel binning turned
      on (MESS:FPU_BIN = 1). The models estimates the dark level
      Dk(x,y,t,T) as a function of column position x, row position y,
      exposure time t in units of milliseconds (as indicated by the
      keyword MESS:EXPOSURE or EXPOSURE_DURATION), and CCD temperature
      T (as indicated by the keyword MESS:CCD_TEMP).
 
      (3) Frame Transfer Smear Correction
      ===================================
 
      Accumulation of signal continues during the finite duration
      of frame transfer induces a streak or frame-transfer smear in the
      wake of an illuminated object in the field of view, parallel to
      the direction of frame transfer. This smear is approximated as:
 
      For each pixel in column x and row y of an image, a correction is
      applied where:
 
      DN_dark_smear(x,y,t,b,f) = DN_dark(x,y,t,b,f) - Sm(x,y,t,b,f)
 
      where
 
      DN_dark_smear(x,y,t,b,f) is dark- and smear- corrected DN,
 
      DN_dark(x,y,t,b,f) is dark-corrected DN,
 
      Sm(x,y,t,b,f) is the estimated smear, and
 
      t is exposure time measured in milliseconds.
 
      (4) Correction for CCD non-linearity
      ====================================
      To remove effects of nonlinearity in WAC image data, each data
      value processed to this point is divided by a function that
      approximates the slight increase in detector sensitivity with
      increasing brightness level. The general form is:
 
      DN_lin = DN_dark_smear/[C1 * Ln(DN_dark_smear) + C2]
      where
 
      DN_dark_smear is the input dark- and smear-corrected DN,
 
      DN_lin is linearized dark- and smear-corrected DN, and
 
      C1 and C2 are constants.
 
      (5) Flat-field correction
      =========================
      The flat field correction removes pixel to pixel differences in
      detector responsivity, so that the responsivity coefficients can
      be expressed as scalars for each filter. There is a separate
      flat-field image for MDIS-WAC and MDIS-NAC (as indicated by the
      keyword INSTRUMENT_ID), without pixel binning turned on
      (MESS:FPU_BIN = 0) or with pixel binning turned on
      (MESS:FPU_BIN = 1), for each separate filter (as indicated by
      the keyword FILTER_NUMBER).
 
      For each pixel in column x and row y of an image, application of
      the correction is
 
      DN_flat(x,y,f,b) = DN_lin(x,y,f,b) / Flat(x,y,f,b)
 
      where
 
      DN_flat(x,y,f,b) is flat-fielded, linearized, dark- and
      smear-corrected DN,
 
      DN_lin(x,y,f,b) is linearized dark- and smear-corrected DN, and
 
      Flat(x,y,f,b) is the value in the appropriate flat-field image.
 
      (6) Conversion from DNs to radiance
      ===================================
      The value that relates corrected DN's measured per unit time to
      radiance is the responsivity, modeled as a function of which camera
      is being used (MDIS-WAC or MDIS-NAC as indicated by the keyword
      INSTRUMENT_ID), its binning state (as indicated by the keyword
      MESS:FPU_BIN), and in the case of the WAC the filter number (as
      indicated by the keyword FILTER_NUMBER). The value also depends on
      CCD temperature (as indicated by the keyword MESS:CCD_TEMP).
 
      To apply responsivity to obtain radiance L, the expression is
 
      L(f) = DN_flat(f) / (t * Resp(f,T,b))
 
      where
 
      L(f) is radiance in filter in units of W / (m**2 microns**1 sr**1),
 
      DN_flat(f) is dark-, smear-, linearity-, and flat field-corrected DN,
 
      t is the exposure time in seconds, and
 
      Resp(f,T,b) is the responsivity in filter f at CCD temperature T and
      binning state b.
 
      Images calibrated to units of radiance have a data type in the
      file name denoted as 'RA'.
 
      (7) Conversion from radiance to I/F
      ===================================
 
      To convert from radiance to I/F (also known as radiance factor, the
      ratio of measured radiance to that which would be measured from a
      white perfectly Lambertian surface), the following is applied:
 
      I_over_F(f, MET) = L(f) / Correct(f,MET)* pi *
       (SOLAR_DISTANCE/149597870.691)**2 / F(f)
 
      where
 
      Correct(f,MET) is the responsivity correction in filter f at time MET,
      which is a real number not equal to unity in WAC images where the
      correction is applied, or unity in other images,
 
      L(f,MET) is calibrated radiance calculated as described above for some
      filter f at time MET,
 
      SOLAR_DISTANCE is that value for distance of the target object from
      the center of the sun in kilometers (as indicated by the keyword
      SOLAR_DISTANCE),
 
      149597870.691 is the number of kilometers in 1 AU, and
 
      F(f) is effective average solar irradiance sampled under the filter
      bandpass.
 
      Images calibrated to units of I/F have a data type in the file name
      denoted as 'IF' if the Correct term is applied. Alternatively the
      Correct term may be omitted in which case the data type is denoted
      as 'IU'.
 
      (8) Treatment of special pixels
      ===============================
 
      Two types of pixels cannot be validly calibrated to either radiance
      or I/F:
 
      (a) Pixels under the dark mask at the edge of the detector do not
      measure light from the scene, yet deviation of their calibrated
      value from zero is a valuable measure of calibration residuals.
      The average calibrated value under the dark mask is reported in the
      label as DARK_STRIP_MEAN, but the actual pixel values are replaced
      by the value indicated in the label for CORE_NULL.
 
      (b) Saturated pixels do not have a known correspondence to scene
      radiance. The pixel values in saturated pixels are replaced by the
      value indicated in the label for CORE_HIGH_INSTR_SATURATION.
 
 
    Data
    ====
      There are up to two data types associated with this volume,
      single-frame calibrated images in units of radiance, or
      single-frame calibrated images in unitless I/F. All data are stored
      as 32-bit PC_REAL.
 
 
    Ancillary Data
    ==============
 
      There are two types of ancillary data provided with this
      dataset:
 
      1. The GEOMETRY directory contains the file GEOMINFO.TXT that points
      to and describes the function of each SPICE kernel relevant to MDIS.
 
      2. The CALIB directory contains a summary of the processing
      required to convert raw data to units of radiance or I/F, as well
      as all of the matrices and coefficients needed. See CALINFO.TXT
      in that directory for more details.
 
 
    Coordinate System
    =================
 
      The cartographic coordinate system used for the MDIS data products
      conforms to the J2000 celestial reference frame for star imaging,
      and the IAU planetocentric system with East longitudes being
      positive for planetary surfaces. A Mercury radius of 2440.0 km was
      used for data products delivered prior to delivery 15. In delivery
      15, that value is updated to 2439.4 km.
 
 
    Media/Format
    ============
 
      The MDIS archive is organized and stored in the directory
      structure described in the Mercury Dual Imaging System (MDIS)
      Calibrated Data Record (CDR) and Reduced Data Record (RDR)
      Software Interface Specification (SIS). The contents of the
      archive, along with fiduciary checksums, are compressed into
      a single 'zip archive' file for transmittal to the PDS Imaging
      node.  The zip archive preserves the directory structure
      internally so that when it is decompressed the original
      directory structure is recreated at the PDS Imaging node.
      The zip archive is transmitted to the PDS Imaging node via
      FTP to the URL specified by the node for receiving it.

Confidence Level Overview
    =========================
      This is a calibrated data set.  Known issues of concern are
      described below.
 
 
    Review
    ======
      This archival data set was examined by a peer review panel
      prior to its acceptance by the Planetary Data System (PDS).  The
      peer review was conducted in accordance with PDS procedures.
 
 
    Data Coverage and Quality
    =========================
 
      The majority of raw EDR data are calibrated to CDRs. Briefly,
      for calibration to radiance, the following criteria are met:
 
      (a) The data represent a scene and not the instrument test
      pattern, as indicated by byte 0 of the data quality index (DQI)
      for the EDR.
 
      (b) The exposure time is greater and zero (zero exposures
      occur in some images due to software features), as indicated
      by DQI byte 1.
 
      (c) Less than 20 percent of the image is saturated (empirically
      this is a threshold dividing wholly corrupted images from
      everything else)
 
      An additional criterion is met for calibration to I/F:
 
      (d) The target of the image is MERCURY, VENUS, EARTH, MOON,
      or CAL_TARGET.
 
      THE USER OF CALIBRATED MDIS DATA IS URGED TO EXAMINE THE
      DQI IN THE LABEL FOR POSSIBLE ISSUES OF DATA QUALITY, AND
      TO UNDERSTAND SOURCES OF UNCERTAINTY IN THE DATA NUMBERED
      (1) THROUGH (14) BELOW. These issues are itemized in order
      of their recognition in the data set.
 
      The 16-byte Data Quality Index or DQI is used to encode
      figures-of-merit into one parameter, including automated
      assessments of validity of the exposure time, presence of an
      excessive number of pixels at or approaching saturation, validity
      of the reported pivot position, quality of spacecraft attitude
      knowledge from the MESSENGER star cameras, CCD temperature
      within range that supports nominal image calibration accuracy,
      and completeness of data within the commanded selection of
      subframes or full frame.
 
      A '1' in any of the fields of the data quality index indicates a
      condition that could adversely affect data quality. A value of '1'
      in bytes 0, 1, or 4 leads to a raw image or EDR not being
      calibrated to a CDR, hence no CDR should exist. A value of
      '1' in byte 2 indicates that radiance values derived from the
      columns of the image experiencing saturation in any row are either
      invalid or suspect. A value of '1' in byte 3 or 5 means
      that even if the image is radiometrically accurate, its
      reconstructed pointing is suspect. A value of '1' in byte 6 (CCD
      out of temperature range at which radiometric calibration is well-
      constrained) is a warning and does not necessarily indicate invalid
      data.
 
      Byte 0: Image source is CCD.
        1 = Image source is test pattern as indicated by
            MESS:SOURCE=1=Test pattern or
            MESS:SOURCE=2=Inverted test pattern.
        0 = Image source is CCD as indicated by MESS:SOURCE=0=CCD.
 
      Byte 1: Valid exposure time.
        1 = Exposure time in ms as indicated by MESS:EXPOSURE equals 0 ms
            (during cruise) or is less than or equal to 2 ms (orbit).
        0 = Exposure time in ms as indicated by MESS:EXPOSURE is greater
            than or equal to minimum valid value.
 
      Byte 2: Presence of an excessive number of pixels at or approaching
      saturation.
 
      As saturation is approached responsivity decreases, and signal
      becomes nonlinear with brightness for small sources. Saturation can
      be exceeded for very bright or large sources once pixel
      antiblooming is overwhelmed. The raw 12-bit DN level indicative of the
      onset of saturation varies between the two CCDs. In the WAC
      (MESS:IMAGER=0) it is approximately 3600; in the NAC
      (MESS:IMAGER=1) it is approximately 3400. If a LUT has been used
      to convert 12-bit to 8-bit DN, then an 8-bit DN value of 255
      also indicates saturation. An 8-bit 255 is encountered before
      saturation of the 12-bit DN in the case of LUT 1. In autoexposure
      mode, the typical threshold for the allowable number of saturated
      pixels is 5 pixels. In manual exposure mode the number of saturated
      pixels is uncontrolled.
 
        1 = There are  > 5 pixels exceeding the DN indicating onset
            of saturation.
        0 = There are  < 5 pixels exceeding the DN indicating onset
            of saturation.
 
      Byte 3: Valid pivot position.
        1 = Pivot position not valid, as indicated by pivot position
            validity flag MESS:PIV_PV=0=invalid.
        0 = Pivot position valid as indicated by MESS:PIV_PV=1=valid.
 
      Byte 4: Filter wheel in position (WAC only; requires
      MESS:IMAGER=0, or else value of this byte = 0).
        1 = Filter wheel not in position, as indicated by either of
            two conditions:
            (a) filter wheel position validity flag MESS:FW_PV=0=invalid,
            (b) an excessive difference between filter wheel resolver
                goal and actual position as given in table below.
        0 = Filter wheel in position as indicated by an allowable
            difference between goal and position, and by MESS:FW_PV=1.
 
            Filter wheel encoder positions
            FILTER_NUMBER   MESS:FW_GOAL   Allowable (abs(MESS:FW_POS -
                                           MESS:FW_GOAL))
            1               17376          +/- 500
            2               11976          +/- 500
            3               6492           +/- 500
            4               1108           +/- 500
            5               61104          +/- 500
            6               55684          +/- 500
            7               50148          +/- 500
            8               44760          +/- 500
            9               39256          +/- 500
            10              33796          +/- 500
            11              28252          +/- 500
            12              22852          +/- 500
 
      Byte 5: Quality of spacecraft attitude knowledge.
        1 = Spacecraft attitude knowledge is bad (MESS:ATT_FLAG is in
            the range 0-3).
        0 = Spacecraft attitude knowledge is good (MESS:ATT_FLAG is in
            the range 5-7).
 
      Byte 6: CCD temperature range.
        1 = CCD out of temperature range at which performance is well
            calibrated (MESS:CCD_TEMP is outside a range of between
            1005 and 1130, which for the WAC is -45C to -11 C, and for
            the NAC is -48C to -14C).
        0 = CCD within well calibrated temperature range (MESS:CCD_TEMP
            is within the stated range).
 
      Byte 7: Completeness of data within the commanded selection of
      subframes or full frame. Missing frames or portions of frames are
      indicated in an EDR with a value of 0 (this cannot be a valid data
      value).
        1 = There are missing data (some pixels populated with 0).
        0 = There are no missing data.
 
      Bytes 8-15: spare.
 
      In addition, the following caveats are applicable to radiance, I/F,
      and map-projected products derived from the EDRs.
 
      (1) WAC CLEAR FILTER. Filter 2 on the wide-angle camera is broad-
      band and designed for star imaging. Even extremely short exposure
      times saturate on Mercury or other typical extended sources.
      Flat-field and responsivity corrections for WAC filter 2
      are less accurate than in other filters.
 
      (2) NAC PSF. Due to mass constraints, the NAC aperture is smaller
      than what is required for diffraction-limited performance. The
      expected size of the Airy disk (approximately, the full-width at
      half-maximum of the point-spread function including only effects of
      diffraction) is > 2 pixels. In practice the PSF is further
      broadened by surface imperfections of optical elements and scatter
      centers on optical surfaces.
 
      (3) COMPRESSION ARTIFACTS. Wavelet compression applied to science
      images is lossy. At higher compression ratios, compression artifacts
      will degrade data precision over spatial scales comparable to or
      smaller than several pixels. The degradation can be greater
      proportionally to the image dynamic range of brightness, if the data
      are converted from 12 to 8 bits in such a way that a 1 DN error
      occupies a greater fraction of the digital dynamic range. Wavelet
      compression was used minimally prior to Mercury orbit. The initial
      configuration in Mercury orbit was to perform 12 to 8 bit conversion
      using LUT0 for the WAC and LUT2 for the NAC, with a wavelet
      compression ratio usually set to 4:1 for color, 8:1 for monochrome
      imaging, or lossless for star imaging. Initial images
      exhibited unexpectedly visible compression artifacts. Beginning
      19 April 2011, LUT0 and LUT2 were replaced with LUT1 which better
      preserves image dynamic range. Beginning 19 May 2011, targeted color
      images began to be acquired with lossless compression. Beginning
      31 May 2011, the wavelet compression ratio for color images was
      reduced to 3:1 for global mapping. Wavelet compression is bypassed
      for a variety of types of images depending on downlink bandwidth and
      availability of space on the spacecraft solid-state recorder.
 
      (4) FRAME TRANSFER SMEAR. At very short exposure times (<7 ms), the
      time for frame transfer is close to the total exposure so that the
      correction for frame transfer smear may leave perceptible
      artifacts. In October 2011 it was discovered that a software
      error in the image calibration pipeline leads to small filter- and
      exposure time-dependent systematic errors in correction of
      frame transfer smear. These are corrected beginning with version 4
      CDRs.
 
      (5) TIME VARIATIONS IN MDIS ATTITUDE. The orientation of MDIS
      relative to the spacecraft reference frame was determined inflight
      using star calibrations to solve for WAC-NAC coalignment, the
      orientation of the pivot plane, and the origin of the reported
      pivot position within the plane. These alignments can be
      affected by thermal state of the spacecraft, or by any other
      events that potentially shift the position of the MDIS base
      relative to the spacecraft star cameras that generate the attitude
      measurements. Mercury and Venus flybys are thermally benign.
      However in Mercury orbit there are thermal perturbations from which
      errors in reported MDIS attitude of up to 350 microradians might be
      expected. Regular star calibrations are conducted in orbit and
      indicate up to two or more types of time variation in attitude of an
      image relative to a given reported pivot position: (a) A small number
      of abrupt shifts of WAC and NAC image alignment, near the time of
      Mercury orbit insertion and again in July 2012, several hundred
      microradians in magnitude. The first is being addressed in an updated
      frames kernel, and was first describe in a frames kernel released in
      conjunction with release 9. The frames kernel in conjunction with
      release 11 and the first release of DDRs also includes the shift of
      NAC image alignment near the time of orbit insertion.
      (b) In addition to abrupt shifts of WAC and NAC image orientation,
      subsequent gradual drift is observed in both from star calibration
      images. That shift and the abrupt one in July 2012 are addressed by
      modifications to the pivot C kernel that incorporate drift
      as part of the pivot attitude. Limited star calibrations that
      determine the relative alignments of the WAC and NAC during the
      orbital phase of the mission indicate that the two fields of view
      drift together, so a single modified pivot C kernel
      is used to describe the pointing for both cameras.
 
      (6) ARTIFACTS IN GROUND-DERIVED FLAT-FIELD CORRECTIONS.
      Two factors make the ground-derived flat-fields less than ideal.
      First, MDIS's structure generated reflections so that in
      the calibration chamber, the illuminated source created glint off
      MDIS blanketing that reflected off the chamber window, adding
      spatially non-uniform stray light to the measurements.
      Eliminating the backscattering off the chamber window required
      acquisition of flat-fields at ambient (room-temperature) conditions
      at which residuals from the dark current correction introduce
      artifacts. Ultimately the latter approach was chosen for ground
      derivations. Second, there are dust donuts (shadows behind out-
      of-focus dust on the CCD cover glass) in the WAC. The locations of
      some WAC dust donuts moved during launch, and other dust donuts
      disappeared. Application of the ground-derived flat-fields
      systematically under- or over-corrects the non-identical WAC dust
      donuts.
 
      The flat-fields were rederived inflight using images of the onboard
      calibration target, and validated with bland areas of Venus during
      Venus flyby 2. These images were taken with a cold CCD, eliminating
      residuals from dark current, and are thought to have WAC dust
      donuts in their 'final' post-launch position. The onboard target
      is uniform, but at the several-percent level shows evidence for low
      spatial frequency glint off the MDIS structure. Therefore to
      rederive the flat field for each WAC filter, an average of several
      images of the calibration target was divided by a median-filtered
      version of the same image, and multiplied by a median-filtered
      ground-derived flat field taken through the same filter with
      the same instrument binning. For the NAC the same procedure was
      used, with the inputs being Venus images.
 
      The following regions of different flat-fields were initially subject
      to errors tracing back to the ground calibration. In calibrated images,
      too-high values of the flat fields yielded too-low values of
      radiance or I/F, or vice versa. The most significant issues were in
      the upper rows of WAC filters 3 and 6.
 
        WAC filter 2: Flat-field values are place-holders only.
 
        WAC not-binned, all filters: In the last 25 columns on the right,
        values that are too high by 1-2%.
 
        WAC not-binned, filter 3: In the top 32 rows, long
        spatial frequency errors of up to a few percent, increasing to
        the top of the image.
 
        WAC not-binned, filter 6: In the top 145 rows, long
        spatial frequency errors of up to a few percent, increasing to
        the top of the image.
 
        WAC binned, all filters: In the last 15 columns on the right,
        values that are too high by 1-2%.
 
        WAC binned, filter 6: In the top 24 rows, long spatial
        frequency errors of up to a few percent, increasing to the top
        of the image.
 
        WAC, filter 5: A weak horizontal banding in the flat-field
        image probably originating from ground calibration.
 
        NAC binned: The most recent flat-field is ground-derived and
        flight measurements do not yield any improvement.
 
      For WAC filters 3 and 6 only, the flat-fields were updated
      in Mercury orbit, to version 5. Approximately 400 images in
      each filter and with DPU binning on and off (4 data sets) were
      calibrated using the previous flat field version, and
      photometrically corrected to remove cross-scene illumination
      gradients. To produce updated flat-fields, the median of the 400
      images was scaled to unity over the same central region as with
      ground-derived flat-fields.
 
      A final update to version 6 in the end-of-mission delivery 15
      used a similar procedure, but more images for improved
      statistics, and excluded images with compression artifacts.
      An improved photometric correction reduced low spatial frequency
      errors. The following filter and binning combinations were
      upgraded to version 6: WAC filters 3 and 6, and binned images
      in WAC filters 3, 4, 5, 6, 7, 9, and 12.
 
      (7) RADIOMETRIC ACCURACY. The responsivities used to convert DNs
      to radiance are based on ground calibrations that were validated
      by comparison of MDIS with MASCS-VIRS measurements of Venus and
      Mercury. Except for 3 of the 12 filters, WAC radiances based on
      ground calibrations yielded a similar spectral shape with
      a 10-15% difference in absolute value. Filters 3 and 6 (at 480
      and 433 nm) were systematically too low and too high
      respectively, so the responsivites used to calibrate them were
      adjusted empirically to improve correspondence with MASCS-VIRS.
      In addition WAC filter 2 (clear filter) radiances have not been
      validated. The relative accuracy of NAC and WAC filter 7 data
      (which correspond in central wavelength) were examined by
      comparing nearly simultaneous images taken during the Mercury 2
      flyby and the NAC calibration was adjusted empirically to
      produce agreement.
 
      Both the NAC and WAC have CCDs whose responsivity to different
      wavelengths of light varies with CCD temperature. Inflight
      including Mercury orbit the CCD operating temperature range is
      typically -10C to -45C; however ground calibration
      measurements were acquired only at +23C, -31C, and -34C
      rendering the initial characterization of the temperature
      dependence inaccurate especially below -34C. M1 and M2
      measurements of comparable surfaces, acquired at CCD
      temperatures of -34C to -43C, were used to improve the
      calibration of CCD temperature dependence over the lower
      end of the temperature range, and this correction was applied
      to version 3 CDRs.
 
      In version 4 CDRs, an additional update to responsivity
      improved temperature dependence over the full operating range.
      The correction was derived empirically by fitting as a function
      of CCD temperature the median values of images acquired at a
      wide range of temperatures but a narrow range of photometric
      geometries.
 
      In version 5 CDRs, a similar correction to responsivity used
      all Mercury images satisfying the illumination criteria.
 
      (8) SCATTERED LIGHT. In the NAC, scattered light from out-of-field
      sources is an issue. The geometry contributing most of the scatter
      is 1-2 fields-of-view sunward of the NAC boresight. For a very
      large, evenly illuminated source that overfills the field-of-view
      by a factor of several, ray-trace studies supported by testing
      during Venus flyby 2 suggest that 2-7% of the radiance measured in
      the field-of-view will have come from out-of-field sources. The
      spatial pattern of the scatter is variable, due to diffuse
      reflections off the internal instrument housing.
 
      The WAC is subject to scattered light originating from within the
      field-of-view or just outside it. In overexposed images, the source
      is evidently multiple reflections off of 13 optical surfaces (2
      sides of each of 4 lenses, the spectral filter, and the CCD cover
      glass, as well as the CCD surface itself). The scatter becomes
      worse at longer wavelengths. Just off the limb of a large
      extended source near 1 field-of-view in size, like Venus or
      Mercury, measured radiance increases with wavelength from 2% to 7%
      of the value measured on the extended source. The value decreases
      with distance off the target more quickly at longer than at
      shorter wavelengths, but remains at 1% hundreds of pixels from
      the source. Conversely, light must be scattering from bright
      parts of an image to dark parts of an image. Averaged over
      sources tens of pixels in area, and away from abrupt brightness
      contrasts, scattered light affects shapes of spectra measured
      from WAC data at least at the 1-2% level, worse near brightness
      boundaries or for small, bright crater ejecta. The expected effect
      is enhanced brightness at >650 nm in dark areas, and decreased
      brightness at >650 nm in small bright areas.
 
      In the end-of-mission delivery 15, a forward model of the
      expected WAC scatter from a given scene was derived using
      optical design software modeling CCD structure and hardware,
      with magnitudes of scatter calibrated against flight measurements.
      The ray trace analysis reveal an in-scene component from light
      diffracted by the CCD and reflected by the CCD cover glass, and
      an out-of-scene component from light reflected off metallic
      surfaces alongside the CCD and back off the cover glass.
      CDRs with bright craters and other albedo non-uniformities located
      at geometries with the strongest scatter were identified, and
      excluded from the final version of color map products (MDRs,
      MD3s, and MP5s).
 
      (9) POINTING UNCERTAINTY. MDIS pivot C-kernels generated prior
      to September 2009 were based on counting steps of the pivot motor.
      There is periodicity at the scale of 3 degrees of pivot rotation
      in the relationship between pivot step size and physical angle,
      leading to reported pivot geometries being in error by up to 350
      microradians in early versions of the pivot C kernel. Inflight
      tests during 2009 supported development of an alternative
      calculation of pivot angle using the pivot position resolver;
      this approach is included in MDIS pivot C kernels since MDIS
      release 5 (after Mercury flyby 3). The uncertainties
      in pivot attitude in latter kernels are of the order of
      35 microradians.
 
      Uncertainty in spacecraft pointing is by requirement less than 350
      microradians, but unofficial estimates from the guidance and
      control team suggest a more typical value near 70 microradians.
      When convolved with error in pivot pointing, this results in the
      majority of error in knowledge of image pointing (exclusive of
      thermal distortion or time variations that can be calibrated out).
 
      An additional and generally larger source of error in map
      projection of MDIS images acquired during Mercury orbit may be
      uncertainty in spacecraft position. At low altitudes that source
      of error dominates, whereas at high altitudes pointing error
      dominates, such that in either case typical errors in map
      projection from the best reconstructed spacecraft position are
      expected to be smaller than 1 km.
 
      In the end-of-mission delivery 15 of SPICE files and map products,
      these errors are reduced by controlling images using c-smithed
      kernels and a global digital elevation model (DEM), both
      derived using a least-squares bundle adjustment of common
      features, measured as tie point coordinates in overlapping NAC and
      WAC-G filter images of Mercury at favorable solar incidence and
      emission angles.
 
      (10) OUT OF FOCUS PARTICULATES IN LONG-EXPOSURE IMAGES OF DEEP
      SPACE. A number of images collected by the MDIS wide-angle camera
      using long (longer than 5 s) exposure times contain artifacts that
      are most likely caused by glint from microscopic particles that
      are in close proximity to the spacecraft. Because the particles
      are out of focus they appear as extended sources. The length of
      these artifacts is most likely due to motion of the particles
      over the significantly large exposure time of the images.
 
      (11) ANOTHER FLAT-FIELD ARTIFACT. In Mercury orbit, it was
      discovered that the flat-field correction in the image column
      immediately adjacent to the 4 column wide dark strip at the edge
      of the CCD is made problematic by low raw DN levels. Therefore
      beginning with version 3 CDRs, in not-binned 1024x1024 pixel
      images, during the calibration process the leftmost 5 columns are
      assigned null values; in 512x512 binned images the leftmost 3
      columns are assigned null values; and in 256x256 binned images
      the leftmost 2 columns are assigned null values.
 
      (12) TIME-VARIABLE WAC RESPONSIVITY. During Mercury orbit it was
      recognized that filter-dependent changes in WAC responsivity on
      the order of +/- 15% occurred over timescales as short as several
      days. Because those variations were not consistent from filter to
      filter, they led to spurious spectral features, which were
      particularly conspicuous near 750 nm. The cause(s) of these
      variations in responsivity are not known, but they could include
      transient radiation effects on the detector or electronics, aging
      of filters, periodic deposition and burn-off of contaminants on
      filters, or incorrect recording of exposure time. An initial
      empirical correction for images acquired in the first year of
      operations was developed at utilized in version 4 WAC CDRs.
 
      For version 5 WAC CDRs in delivery 15 at end of mission, an
      updated correction covers the full duration of the orbital
      phase. Overlaps between color image sets in color mapping campaigns
      were used to derive a multiplicative correction factor for each
      filter and for each Earth day (2-3 orbits). It used all image data
      with a pixel scale >50 m and incidence angle (measured at the
      center pixel of the 750-nm image in the image set) of <80 degrees.
      The images were calibrated to I/F using the standard calibration
      with Correct(f,MET) set to unity, the latest
      version of temperature-dependence of detector responsivity, the
      latest correction for frame-transfer smear, and newly derived
      flat-field corrections. Images were then photometrically
      normalized to 30 degrees incidence angle, 0 degrees emission angle,
      and 30 degrees phase angle. Any portions of image sets lacking
      coverage in all filters, or with incidence angles >70 degrees or
      emission angles >30 degrees, were trimmed. Image sets were mapped
      to an equal-area (sinusoidal) projection at 4 km/pixel. All images
      acquired on each day were mosaicked for each filter, typically
      resulting in a fairly narrow north-south strip that had substantial
      overlap with surrounding days. With this dataset, multiplicative
      correction factors for each day were calculated through a weighted
      least-squares optimization that minimized the discrepancy between
      the median values for all spatial overlaps. The optimization was
      performed in two steps. First, a mosaic of data acquired before
      22 May 2011 was held as constant, because this dataset was seen to
      be largely self-consistent. However, these data covered only a
      fraction of the planet, so a mosaic with greater coverage was
      created from images that were corrected in the first iteration with
      low residual values. In the second step, all data were allowed to
      vary in the simultaneous optimization, with the new mosaic held as
      reference. Correction factors were derived for days on which no data
      were included in our optimization process by interpolating between
      adjacent days. Filters with center wavelengths at 700, 950, and
      1020 nm were not used for regional or global mapping and thus did not
      have enough overlap to derive correction factors with this method.
      Instead, their empirical correction was derived by comparing them
      with a synthetic global mosaic created by linear interpolation from
      images acquired with adjacent filters.
 
      Final 1-, 3-, 5-, and 8-color mosaics use CDRs with this updated
      correction. An analysis of overlap among individual images shows
      that residual differences (which include errors from calibration,
      scattered light, and possible incomplete correction of photometric
      variation) average <2% for the majority of the planet.
 
      (13) TEMPERATURE-DEPENDENCE OF WAC and NAC FOCAL LENGTH. The
      star calibrations used to track the position of the MDIS pivot
      base, augmented by special calibrations using fields with high
      spatial densities of stars, sample much of the range of thermal
      environments experienced in orbit. Analyses of these data showed
      that focal length of each camera is correlated with temperature
      of the focal plane housing 'FOCAL_PLANE_TEMPERATURE', and that
      focal length varies over the cameras' operating temperature range
      by several parts in 10,000. This dependence is included in an
      updated delivery of the instrument kernel associated with PDS
      release 11.
 
 
    Limitations
    ===========
      None