Orthogonal processing of multispectral images - Sensitivity of the Karhunen-Loeve coefficients to sampling
Abstract
The Karhunen-Loeve orthogonal expansion is a useful tool which enables decorrelation of the spectral channels and calculation of principal linear combinations. In the present paper, the effects of noise are studied by analyzing the effect of sampling on the images, which amounts to analyzing the effect of resolution on the results of Karhunen-Loeve analysis. The analysis enables one to predict how many principal coefficients (images) must be retained as a function of the real multispectral information for a given value of ground resolution. It is shown that the errors due to sampling noise can change the order of certain eigenvalues and even their number, that is, the dimensions of the multispectral image representation.
- Publication:
-
5th Conference on Space Optics
- Pub Date:
- 1976
- Bibcode:
- 1976spop.conf..137L
- Keywords:
-
- Data Processing;
- Data Sampling;
- Imaging Techniques;
- Karhunen-Loeve Expansion;
- Multispectral Photography;
- Digital Data;
- Error Analysis;
- Iterative Solution;
- Linear Transformations;
- Numerical Stability;
- Orthogonal Functions;
- Random Processes;
- Signal To Noise Ratios;
- Statistical Analysis;
- Instrumentation and Photography