A Study of Texture Analysis Algorithms.
Abstract
This research has focused upon developing improved texture analysis algorithms. Work performed during the second year of the grant has shown that the Spatial Gray Level Dependence (SGLDM) texture analysis algorithm is a superior algorithm under fairly weak assumptions. For this reason our subsequent work has continued the development of the SGLDM method. Tiling theory has been combined with the SGLDM analysis procedure to create a structural (SSA) analyzer for texture patterns. Recent work has focused upon determining measures derived from the SGLDM cooccurence matrices that characterize texture patterns. It has been shown that the commonly used measures are inadequate. A texture generation procedure has been developed and this has been used to generate new measures based upon the perceptual concepts of uniformity and proximity. These measures offer promise of developing measures related to perceptual features. Experiments were also conducted which shows that the SGLDM algorithm can discriminate known counterexamples to the Julesz conjecture. Thus the robustness of the SGLDM has been further established over this theoretically troublesome class of textures. (Author)
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 24, 1981
- Accession Number
- ADA130034
Entities
People
- Charles A. Harlow
- Richard W. Conners
Organizations
- Louisiana State University