Document Actions
Minimum Noise Fraction
Minimum Noise Fraction (MNF) analysis identifies the locations of spectral signature anomalies. This process is of interest to explorationists because spectral anomalies are often indicative of alterations due to hydrocarbon microseepage.
MDA can perform this analysis on any multispectral data set. Our geologists then interpret the MNF image using information on known hydrocarbon accumulations in the region to identify anomalous areas that may be due to microseepage.
How It Works
The minimum noise fraction (MNF) transform is used to determine the inherent dimensionality of image data to segregate noise in the data and to reduce the computational requirements for subsequent processing (see Boardman and Kruse, 1994). This method is similar to principal component (PC) analyses that have been used for a long time in multispectral image processing, but involves an extra preceding step.
MDA uses a MNF algorithm modeled after that created by Green, et al. (1988). The first transformation applied to the data decorrelates and rescales the noise in the data based on an estimated noise covariance matrix calculated for the scene prior to MNF transformation. This results in transformed data in which the noise has unit variance and no band-to-band correlations. Then a standard PC transformation is applied to the noise-whitened data.
The data can be divided into two parts: one part associated with large eigenvalues and coherent eigenimages, and a complementary part with near-unity eigenvalues and noise-dominated images. Subsequent image processing based on the MNF results is improved by the removal of the noise. For a full discussion of PC analysis in image processing, see Hall (1979).
The MNF transformed data can then be used for classification using any standard classification scheme or by manually choosing training classes on scatter plots. However, as the transformed bands are eigenimages, there is no direct correlation between brightness and composition. Therefore, band to band correlations between pixels imply spectral similarity, but the specific material responsible for it cannot be identified directly.
