Aerial image analysis using fractal-based models

The fractal dimension and variants of the fractal dimension have been used to analyze visual textures in images, and the vast majority of research assumes that the fractal model closely approximates the observed image. However, the difference between the image and the assumed fractal model can be significant, and it can provide useful information about image texture. In fact, this fractal error can itself serve as a texture descriptor. A localized version of fractal error is developed and used for image segmentation. The fractal error is examined as a potential texture feature within the context of segmentation of aerial reconnaissance images. Accuracy on the order of 85 percent was demonstrated in three typical examples.<
>