Linearized Shadow and Water Indices
The next set of indices we introduce within the same statistical framework used in the previous chapter are the shadow–water indices (SWI). We benefit from these indices to detect lakes in residential regions in the following chapters. Therefore, they also provide valuable information in analyzing multispectral images. In the Ikonos spectrum, water shows an increasing response curve until the blue band, it reaches the maximum in this region and then decreases monotonically to the near-infrared. So a representative shadow–water index should be composed of high blue values first. Ideally, it should also consider the green and red bands, but the green band also responds strongly to vegetation and this impairs the shadow or water observation. Hence, the index should include blue and red bands at least. To obtain such an index, we applied the same framework we used for theNDVI derivation using principal components analysis with the blue, red, and near-infrared bands. Based on the combinatorial search (and trying to maximize blue and red band coefficients) we obtain the best performing shadow–water index for each dimension.
KeywordsMultispectral Image High Entropy Vegetate Region Blue Band Previous Chapter
- 1.V. Klemas, Remote sensing of estuaries: An overview. IEEE Ocean, pp. 302–309 (1986) Google Scholar
- 5.A.T.S. Ho, inProceedings of IPA (1997), pp. 848–851 Google Scholar
- 8.G. Lisens, P. Kempencers, F. Fierens, J. Van Rensbergen, inProceedings of IGARSS, vol. 2 (2000), pp. 834–836 Google Scholar
- 10.H. Spitzer, R. Franck, M. Kollewe, N. Rega, A. Rothkirch, R. Wiemker, inProceedings of IGARSS, vol. 5 (2001), pp. 2256–2258 Google Scholar
- 11.T.E. Avery, G.L. Berlin,Fundamentals of Remote Sensing and Airphoto Interpretation, 5th edn. (Macmillan Co, New York, 1992) Google Scholar