FRACTAL ANALYSIS OF HYPERSPECTRAL IMAGES

Abstract

Hyperspectral Imaging is simultaneous acquisition of images in many relatively narrow, contiguous, and/or non-contiguous spectral bands throughout the ultraviolet, visible and infrared portions of Electro-Magnetic Spectrum. The value of an hyperspectral imaging (also known Imaging Spectrometry or Spectroscopy) lies in its ability to provide a high resolution reflectance spectrum (in region from 0.4-2.5 μm) for each picture element in the image. These are used to identify a large range of surface cover materials as many surface cover material have diagnostic absorption features that are only 20-40 nm wide. The objective of this study is to analysis hyperspectral images using fractal mathematics. The Spectral Reflectance Curve (SRC) of any land cover represents its characteristics. Thus, Fractal Dimension (FD) of SRC of any pixel location is expected to represent the characteristics of the land cover it contain. There are different methods for estimation of fractal dimension of waveforms, categorized into these types: variance method, spectral method and morphological covering method. In this study, FD for SRC of each pixel in air-borne or space-borne hyperspectral images is estimated to: 1. obtain a feature for classification of hyperspectral images; 2. reduce the dimensionality of hyperspectral images; 3. find the anomaly present in pixel content Variations found in Fractal Dimension Index (FDI), FD , of pixel's SRC, indicate the anomalies present in the pixels belonging to objects present on the surface of earth, and thus may be useful for higher level classification in future. Further, Fractal Dimension Index Image or FDI Image is developed for visual appraisal of the scenes.