Segmentation (Image Processing) - Important Concepts in Signal Processing, Image Processing and Data Compression - page 38

This server allows:

to obtain customized experimental texture mosaics and their
corresponding ground truth,



to obtain the benchmark texture mosaic set with their corresponding
ground truth,



to evaluate your working segmentation results and compare them with the
state of art algorithms,



to include your algorithm (reference, abstract, benchmark results) into the
benchmark database,



to check single mosaics evaluation details (criteria values and resulted
thematic maps),



to rank segmentation algorithms according to the most common
benchmark criteria,



to obtain LaTeX coded resulting criteria tables.



Computer-generated texture mosaics and benchmarks are composed from the following
test image types:

monospectral textures



multispectral textures



BTF (bidirectional texture function) textures BTF Bonn Database



ALI hyperspectral satellite images Earth Observing 1



rotation invariant texture set



scale invariant texture set



illumination invariant texture set 

All generated texture mosaics can be corrupted with additive Gaussian noise, Poisson or
salt&pepper noise.

Hold-out trainee sets are supplied in the supervised mode.

Submitted results are stored and used to update the ranking of algorithms based on a
selected region-based criterion from the following criteria set:

region-based (including the sensitivity graphs)



· US - under-segmentation · ME - missed error · NE - noise error
· CS - correct segmentation · OS - over-segmentation

pixel-wise



· O - omission error · C - commission error,
· CA - class accuracy · CO - recall = correct assignment,
· CC - precision = object accuracy
· I. - type I error · II. - type II error