Feature analysis and reduction of laws texture measure
Rand, Robert S.; Shine, James A.
Research noteAbstract: The Laws Texture Measure was tested in an autoevaluation experiment and compared to some of the texture measures previously studied at ETL. In addition, four methods of component reduction were evaluated. Laws Texture was found to be the most effective measure for identifying a Buildings/Roads class; however, it was less effective on other classes. A technique that transforms the original components into principal components and then arranges the transformed components in an order that maximizes the divergence was selected as the most effective component reduction method.
U.S. Army Engineer Topographic Laboratories.Engineer Research and Development Center (U.S.)
Bayes Classifier; Digital imagery; Divergence; Laws texture measure; Pattern recognition; Principal components
ETL ; 0343.
Approved for public release; distribution is unlimited.