The decorrelation stretch transformation
暂无分享,去创建一个
Abstract The three-stage decorrelation stretch analysis—transformation to principal component scores; contrast-stretching of the scores to equalise the variances; and reversal of the initial transformation—is shown to produce linear combinations of the original bands which are uncorrelated and have unit variances. As such, the effectiveness of the decorrelation stretch analysis in producing enhanced displays depends fortuitously on the particular contrasts which result. For some examples, setting the first scaled principal component to zero before reversing the initial transformation has little effect on the decorrelation stretch coefficients. In other examples, a small change in the eigenvector defining the first principal component results in only a small change in the decorrelation stretch coefficients and yet produces a distinctly different decorrelation stretch image. A more general formulation of the decorrelation stretch analysis involving simplification and/or exclusion of some of the first-stage ...
[1] A. Goetz,et al. Mineralogic Information from a New Airborne Thermal Infrared Multispectral Scanner , 1983, Science.
[2] R. E. Walker,et al. Color enhancement of highly correlated images. I - Decorrelation and HSI contrast stretches. [hue saturation intensity , 1986 .
[3] John A. Richards,et al. Remote Sensing Digital Image Analysis , 1986 .
[4] C. R. Rao,et al. Linear Statistical Inference and its Applications , 1968 .