DocumentCode :
3020310
Title :
Texture analysis using gaussian weighted grey level co-occurrence probabilities
Author :
Jobanputra, R. ; Clausi, D.A.
Author_Institution :
University of Waterloo
fYear :
2004
fDate :
17-19 May 2004
Firstpage :
51
Lastpage :
57
Abstract :
The discrimination of textures is a significant aspect in segmenting SAR sea ice imagery. Texture features calculated from grey level co-occurring probabilities (GLCP) are well accepted and applied in the analysis of many images. When calculating GLCPs, each co-occurring pixel pair within the image window is given a uniform weighting. Although a novel technique, co-occurring texture features have a tendency to misclassify and erode texture boundaries due to the large window sizes needed to capture meaningful statistics. A method is proposed whereby co-occurring pixel pairs closer to the center of the image window are assigned larger cooccurring probabilities according to a Gaussian distribution. By using a Gaussian weighting scheme to calculate the GLCPs, less significance is given to pixel pairs that are on the outlying regions of the window, which have a tendency to produce erroneous statistics as the image window overlaps a texture boundary. This method proves to preserve the edge strength between textures and provides better segmentation at the expense of computational complexity.
Keywords :
Image analysis; Image segmentation; Image texture analysis; Noise robustness; Pixel; Probability; Remote monitoring; Sea ice; Statistical distributions; Synthetic aperture radar;
fLanguage :
English
Publisher :
ieee
Conference_Titel :
Computer and Robot Vision, 2004. Proceedings. First Canadian Conference on
Conference_Location :
London, ON, Canada
Print_ISBN :
0-7695-2127-4
Type :
conf
DOI :
10.1109/CCCRV.2004.1301421
Filename :
1301421
Link To Document :
بازگشت