Contents || Index
04/20/2005

Outliers

Out of range data

Often images or data cubes have pixels with values that are way too large, which render most of the image as featureless black.  There are several ways to take care of this problem.

For Images (in the Lispix Menu Bar):   (Help / Samples & Info / Images / Rutile 16 bit is a good sample image to use.  Example using the Rutile image.)

  1. Scale / trimmed  Useful for ignoring the brightest and darkest pixels in an image.  You type in the percent on each end of the distribution to ignore.
  2. Scale / fixed limits Same result as trimmed scaling, except you put in the brightness level range to be used -- any pixels outside this range are ignored.
  3. Scale / equalized A severe sort of enhancement, but it will take care of the outliers and show something in the image.
  4. CLUT / display Gray Level component If you use a slider to bring out detail in an image, the outliers will probably be clipped, along with the image over all intensities being changed.  The gray level component is an 8-bit image of what you see, and can be saved as a TIFF file.

For Data Cubes (in the Cube button of the Data Cube Tool);

  1. Cube / trimmed scaling  Each slide is scaled independently of the others, using the percent that you choose.  This usually makes all of the slides in a cube show something, but note that a spectrum of such a cube will be garbaged.
  2. Cube / clip to fixed limits. All of the slices are scaled to the same limits, clipping outliers.  The spectra from such a cube will be OK, except that peaks will be cut off at the limit, in other words they will have flat tops.