The analysis and quantitative measurement of image texture is a complex and intriguing problem that has recently received a considerable amount of attention from the diverse fields of computer graphics, human vision, biomedical imaging, computer science, and remote sensing. In particular, textural feature quantification and extraction are crucial tasks for each of these disciplines, and as such numerous techniques have been developed in order to effectively segment or classify images based on textures, as well as for synthesizing textures. However, validation and performance analysis of these texture characterization models has been largely qualitative in nature based on conducting visual inspections of synthetic textures in order to judge the degree of similarity to the original sample texture imagery. In this work, four fundamentally different texture modeling algorithms have been implemented as necessary into the Digital Imaging and Remote Sensing Synthetic Image Generation (DIRSIG) model. Two of the models tested are variants of a statistical Z-Score selection model, while the remaining two involve a texture synthesis and a spectral end-member fractional abundance map approach, respectively. A detailed validation and comparative performance analysis of each model was then carried out on several texturally significant regions of two counterpart real and synthetic DIRSIG images which contain differing spatial and spectral resolutions. The quantitative assessment of each model utilized a set of four performance metrics that were derived from spatial Gray Level Co-occurrence Matrix (GLCM) analysis, hyperspectral Signal-to-Clutter Ratio (SCR) measures, mean filter (MF) spatial metrics, and a new concept termed the Spectral Co-Occurrence Matrix (SCM) metric which permits the simultaneous measurement of spatial and spectral texture. These performance measures in combination attempt to determine which texture characterization model best captures the correct statistical and radiometric attributes of the corresponding real image textures in both the spatial and spectral domains. The motivation for this work is to refine our understanding of the complexities of texture phenomena so that an optimal texture characterization model that can accurately account for these complexities can be eventually implemented into a synthetic image generation (SIG) model. Further, conclusions will be drawn regarding which of the existing texture models achieve realistic levels of spatial and spectral clutter, thereby permitting more effective and robust testing of hyperspectral algorithms in synthetic imagery.
Library of Congress Subject Headings
Image processing--Digital techniques; Imaging systems--Image quality; Remote sensing
Imaging Science (MS)
Department, Program, or Center
Chester F. Carlson Center for Imaging Science (COS)
Scanlan, Neil W., "Comparative performance analysis of texture characterization models in DIRSIG" (2003). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus