In this master thesis we propose high order symmetry derivatives of gaussians for texture classification. The symmetry derivative approach is applied to a multiresolutional pyramid structure, which finally results in a more dimensional feature space represented by high order complex moments. For visualization of results the features are presented to a image classification and segmentation algorithm using multidimensional clustering and orientation adaptive boundary refinement. Test images are generated to validate the functionality of symmetry derivatives for textures with multiple orientations and in this context we propose an extension of double angle representati onto visualize multiple local orientations. Further experiments with real texture images are carried out to improve the quality of the feature space byusing different methods for preprocessing and feature space dimensionality reduction.
| Identifer | oai:union.ndltd.org:UPSALLA1/oai:DiVA.org:hh-12002 |
| Date | January 2003 |
| Creators | Pomwenger, Werner |
| Publisher | Högskolan i Halmstad, Sektionen för Informationsvetenskap, Data– och Elektroteknik (IDE) |
| Source Sets | DiVA Archive at Upsalla University |
| Language | English |
| Detected Language | English |
| Type | Student thesis, info:eu-repo/semantics/bachelorThesis, text |
| Format | application/pdf |
| Rights | info:eu-repo/semantics/openAccess |
Page generated in 0.0022 seconds