The pixel2map program projects an ISIS level0 or level1 cube to a map-projected, ISIS level2 cube. Please note that pixels that fall in either polar region (i.e. latitude of 90.0 degrees or -90.0 degrees) cannot currently be processed and are assigned as NULL pixels (i.e. this is currently unsupported).

In order for pixel2map to run successfully, input cubes must have SPICE data. The program spiceinit should be used to attach the appropriate SPICE data to input cubes. If a shape model is provided (or added in spiceinit by default), orthorectification is supported through the determination of the field-of-view (FOV) latitude/longitude coordinates as they map onto the digital elevation map (DEM). The pixel2map program also requires a map projection specification. The map projection is defined using a PVL file specified with the MAP parameter. The maptemplate program can be used to create a map projection file.

The cam2map program also projects a cube and is recommended for most mapping applications. The cam2map program assumes contiguous pixels and will interpolate to fill any gaps that occur in the output image during projection. The pixel2map program will preserve these gaps where the pixels are not contiguous. This program is designed to project cubes for the following cases:

• pixels do not overlap spatially
• data is smeared from long exposure times and drifting (see FOVRANGE=FULLEXPOSURE)
• data is captured from spectrometers
• CRISM
• DAWN VIR
• NEARS MSI
• VIMS

It is important to note that when projecting multiple spectrometer images within a single run of pixel2map, the hyperspectral information present in the input may be compromised in the output projection because individual overlapping pixels are averaged in the output pixel. With a single observation as input, the hyperspectral information will be preserved. With multiple observations as input (by using FROMTYPE=FROMLIST and FROMLIST), each observation may have different geometry depending on the timeframe the observations were acquired. Input pixels from each input observation mapped to an overlapping output surface pixel will be averaged.

The following table indicates how pixel2map and cam2map differ:

pixel2map	cam2map
Uses a forward driven algorithm for projection.	Uses a reverse driven algorithm for projection.
Each input pixel is converted to a FOV polygon, typically using the latitude/longitude of the pixel corners, and rasterized into the output projected image by calculating output pixels that fall in the boundary of the input pixel.	Each input pixel is not converted to a polygon to be rasterized. Up to 16 input pixels are interpolated around the center latitude and longitude of the output map pixel.
The space between pixels is retained in the output projected cube.	The space between pixels are interpolated thus gaps are filled in the output projected cube.
The output cube will be spatially accurate for non-continuous pixels.	The output cube may not be spatially accurate as output pixels are interpolated.

The examples below are Mars Reconnaissance Orbiter/CRISM cubes where scan lines do not overlap in their original state. The major difference is in the output of the application used to project the cube. Both applications will project the cube, but will also result in a different spatial outcome.

Left Cube: cam2map was chosen to project the cube. Notice the scan lines have been interpolated/filled. If you want to preserve gaps, cam2map will not achieve this result (but nearest-neighbor interpolation will be more accurate). In addition, the cube is no longer accurate (traceable for this type of instrument).

Right Cube: pixel2map was chosen to project the cube. Notice the gaps are retained in the output and the cube is spatially correct.

Mapping Parameterization

Several parameters of the map projection have default values when not specified in the map projection file. The default projection is the system sinusoidal projection. The following table indicates how the defaults are established:

PARAMETER	DEFAULT
TargetName	Read from Instrument group in the input cube labels
EquatorialRadius PolarRadius	Read from SPICE PCK file set during spiceinit. The PCK file is defined in the Kernels group via the TargetAttitudeShape keyword
LatitudeType	Planetocentric
LongitudeDirection	PositiveEast
LongitudeDomain	Normally, 360. However, if the 180 domain causes less area to need to be projected then 180.
MinimumLatitude MaximumLatitude MinimumLongitude MaximumLongitude	Computed from the input cube or read from the map file. However, any combination of the four values can then be overridden by the user. The values the user specifies are expected to be in the coordinate system of the projection.
PixelResolution Scale	Computed from the input cube or read from the map file. The value can be overridden by the user.

If you only entered the input cube (FROM) and output cube (TO) and changed no other parameters, the following is the default Mapping group:

  Group = Mapping
    TargetName             = Obtained from the Instrument group
    EquatorialRadius       = Obtained from TargetAttitudeShape kernel
    PolarRadius            = Obtained from TargetAttitudeShape kernel

    LatitudeType           = Planetocentric
    LongitudeDirection     = PositiveEast
    LongitudeDomain        = 360 (Could be automatically adjusted to 180 by LONSEAM)

    MinimumLatitude        = Computed from the input camera cube
    MaximumLatitude        = Computed from the input camera cube
    MinimumLongitude       = Computed from the input camera cube
    MaximumLongitude       = Computed from the input camera cube

    ProjectionName         = Sinusoidal
    CenterLongitude        = Average of MinimumLongitude and MaximumLongitude
    PixelResolution        = Computed from the input camera cube
  EndGroup
    

The map file can be an existing map projected (level2) cube. A level2 cube has PVL labels with a Mapping group that is used to determine the default output map geometric extents and resolution. Depending on the values of the input parameters, the output cube can use some or all of the keyword values of the map file. For instance, setting MATCHMAP=true causes all of the mapping parameters to come from the map file, resulting in an output cube having the same number of lines and samples as the map file. If MATCHMAP=true and the map file is missing a keyword like PixelResolution, the application will fail with a PVL error. Setting MATCHMAP=false allows for some of the mapping components to be overridden by the user or computed from the FROM cube.

If you are attempting to construct a mosaic, it is important that the PixelResolution, EquatorialRadius, PolarRadius, LatitudeType, LongitudeDirection, LongitudeDomain, ProjectionName, and projection specific parameters (e.g., CenterLongitude, CenterLatitude) are the same for all cubes. That is, you should create one map file and use it as input for all the cubes in your mosaic. By letting the minimum and maximum latitude and longitude values default, the application will determine the coverage of each image. However, if the mosaic Latitude and Longitude range is entered, each output image will be projected to the full size of the mosaic resulting in large file sizes and images with many NULL pixels. The following Mapping group could be used for mosaicking:

 Group = Mapping
   ProjectionName         = Sinusoidal
   CenterLongitude        = 0
   PixelResolution        = 100 <meters>
 EndGroup
   

Finally, depending on the projection, problems can occur with cubes that fall on the projection longitude seam. For example, if you are making a mosaic with LongitudeDomain = 360 and your cube crosses the 0/360 seam, this program would compute the default longitude range of the cube to MinimumLongitude = 0 and MaximumLongitude = 360. A very large output image could be created depending on the pixel resolution. The LONSEAM parameter allows you to selectively handle this case. If you are making mosaics near the seam you will need to understand and alter the default for this parameter. Problems at the Longitude Seams of The ISIS Workshop "Learning About Map Projections" includes an example to help illustrate the problem.

Output of pixel2map

A single input file (FROM) produces a projected level2 cube. A list of files (FILELIST) produces an averaged mosaic as output. An additional count cube with one band that contains the number of input pixels averaged into each output pixel is created along with the output cube or mosaic.

Instantaneous FOV (IFOV) vs. Full FOV

As mentioned above, pixel2map uses a forward-driven mapping algorithm to create a projection. In short, this means that pixel2map determines a polygon (footprint) FOV for each input pixel and rasterizes it into the output image space. Depending on what the METHOD parameter has been set to, the input pixel DN value is included in the average of any output pixels that fall within the polygon.

The default behavior of pixel2map is FOVRANGE=INSTANTANEOUS. This means that the IFOV at the center time of the exposure is being used to determine each pixel's footprint. For spectrometers that have longer exposure times, this method does not adequately address any smearing that can occur due to drifting during the exposure. In order to account for smearing, FOVRANGE=FULLEXPOSURE can be used. This method determines a pixel's polygon by finding the IFOV polygons at the start time, middle time, and end time of the exposure and then drawing an envelope around these polygons. The images below visually demonstrate the differences between the two FOVRANGE parameter values.

For more details, see the parameter description for FOVRANGE or examples 3 and 4.

In this example, we demonstrate projecting a CRISM image using the default pixel2map parameters. Note that the output retains the gaps in the data, indicating that the input pixels do not overlap spatially.

In this example, we project a VIMS-IR image using the LONSEAM=CONTINUE parameter setting. This means when we cross the longitude boundary in the image, the longitude domain is not automatically reset. The output image is split at the longitude boundary that it crosses; therefore the entire range of the target body is included in the output, a full 360 degree longitude. Note that a count cube is still created when running the commands below, it is just ignored in this example.

In this example, we project a DAWN VIR image using an instantaneous field of view and a full field of view to determine the output pixels. This demonstrates FOVRANGE=INSTANTANEOUS (default setting for FOVRANGE). Note that a count cube is still created when running the commands below, it is just ignored in this example. Compare this with FOVRANGE=FULLEXPOSURE in Example 4.

In this example, we project a DAWN VIR image and use each pixel's full field of view to determine the output pixels. This example demonstrates FOVRANGE=FULLEXPOSURE. Note that a count cube is still created when running the commands below, it is just ignored in this example.