Paper: | MP-L2.5 |
Session: | Image And Video Segmentation II: Texture Segmentation |
Time: | Monday, September 17, 16:10 - 16:30 |
Presentation: |
Lecture
|
Title: |
STRUCTURAL TEXTURE SEGMENTATION USING AFFINE SYMMETRY |
Authors: |
Heechan Park; University of Warwick | | |
| Graham Martin; University of Warwick | | |
| Abhir Bhalerao; University of Warwick | | |
Abstract: |
Many natural textures comprise structural patterns and show strong self-similarity. We use affine symmetry to address the self-similarity relationship; that is a patch of texture can represent other similar patches after affine transformation. Based on this model, we demonstrate efficient segmentation of structural textures. The algorithm employs the Multi-resolution Fourier Transform and uses Radon analysis to obtain a directional shape from the local spectrum. This process offers reasonable noise robustness for directional feature detection. From the obtained shape, a feature is extracted using the affine-invariant Fourier descriptor. Experimental evaluation on structural textures shows encouraging results and application on natural images demonstrates identification of texture objects. |