EM 1110-2-2907
1 October 2003
f. RMS error for each GCP is recorded in a matrix spreadsheet. A total RMS error of
0.7742 is provided in the upper margin of Figure 5-4a.
Figure 5-4. GCP selection display modules.
pling the image using a "nearest neighbor" re-sampling technique. The software easily
performs this process. Nearest neighbor re-sampling uses the value of the nearest pixel
and extracts the value to the output, or re-sampled pixel. This re-sampling method pre-
serves the digital number value (spectral value) of the original data. Additional re-sam-
pling methods are bilinear interpolation and cubic convolution, which recalculate the
spectral data. The image is projected subsequent to re-sampling, and the file is ready to
be saved with a new name.
Recommendation: Naming altered data files and documenting
procedures
Manipulating the data alters the original data file. It is therefore a good idea
to save data files with different names after performing major alterations to
the data. This practice creates reliable data backup files.
Because of the number of data files an analysis can create, it is best to clearly
name the altered image files with the procedure name performed on the
image (i.e., "TmSept01warped" indicates Thematic Mapper data collected
September 2001, warped by user). Be sure to document your procedures and
parameters used in a journal or a text file. Include the name of the altered
file, changes applied to the data, the date, and other useful information.
5-11
Previous Page
