Name
Abstract | ||
|---|---|---|
Spectral correlation, as quantiøed by the elements of the covariance matrix, plays a prominent role in the development of optimumstatisticalalgorithmsforhyperspectraldataexploitation(1). Indeed,themostusefulstatisticalmodelsforhyperspectral image modeling, namely the multivariate normal distribution and the multivariate t-distribution, are parametrized by the spectral covariance matrix (2). Therefore, it is extremely useful toønd ways to reveal where the important inter-band information lies and how it can be exploited to design effective and eføcient data analysis algorithms. In this paper, we use the theory of optimum linear prediction (3) and the concept of partial correlation coeføcients to provide a simple theoretical framework for understanding where spectral inter-band information is concentrated in hyperspectral data and how this information can be effectively and eføciently exploited. Given that the inverse covariance matrix appears in many detection and classiøcation algorithms, we investigate how spectral information affects its structure and how it may lead to improved algorithms (4). We also derive a spectral-innovations representation of hyperspectral data and we compare it with the widely used principal components transformation (5) as a data whitening tool. All these concepts and algorithms are demonstrated using hyperspectral imaging data from the AVIRIS sensor. To illustrate some of the ideas discussed in this paper, we consider the estimation of the re¿ectance xi in the ith-band using a linear combination of the remaining bands x1 ,...,x i−1 ,x i+1 ,...,x p. The estimation error is |
Year | DOI | Venue |
|---|---|---|
2008 | 10.1109/IGARSS.2008.4779059 | IGARSS |
Keywords | Field | DocType |
ei = xi − ˆ xi = p,partial correlation,hyperspectral sensors,image processing,principal component,reflectivity,data analysis,estimation,hyperspectral imaging,wiener filter,multivariate t distribution,multivariate normal distribution,gaussian distribution,covariance matrix,sparse matrices,correlation | Multivariate t-distribution,Pattern recognition,Multivariate statistics,Image processing,Hyperspectral imaging,Multivariate normal distribution,Artificial intelligence,Statistical model,Covariance matrix,Sparse matrix,Mathematics | Conference |
Citations | PageRank | References |
4 | 0.58 | 1 |
Authors | ||
3 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Dimitris Manolakis | 1 | 31 | 10.82 |
| Ronald B. Lockwood | 2 | 15 | 5.30 |
| Thomas W. Cooley | 3 | 45 | 17.37 |