Open AccessTexture spectral similarity criteria
Author(s) -
Havlíček Michal,
Haindl Michal
Publication year - 2019
Publication title -
iet image processing
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.401
H-Index - 45
eISSN - 1751-9667
pISSN - 1751-9659
DOI - 10.1049/iet-ipr.2019.0250
Subject(s) - hyperspectral imaging , similarity (geometry) , texture (cosmology) , pixel , monotonic function , computer science , artificial intelligence , pattern recognition (psychology) , computer vision , image (mathematics) , mathematics , mathematical analysis
New similarity criteria capable of assessing spectral modelling plausibility of colour, bidirectional texture functions (BTF), and hyperspectral textures are presented. The criteria credibly compare the multi‐spectral pixel values of the textures. They simultaneously consider the pixels of similar values and their mutual ratios. It allows support of the optimal modelling or acquisition setup development by comparing the original data with its synthetic simulations. Analytical applications of the criteria can be spectral based texture retrieval or classification. The suggested criteria together with existing alternatives are extensively tested and compared on a wide variety of colour, BTF, and hyperspectral textures. The performance quality of the criteria is examined in a long series of thousands specially designed monotonically degrading experiments, where proposed ones outperform all tested alternatives.