SPECTRALLY-CONSISTENT RELATIVE RADIOMETRIC NORMALIZATION FOR MULTITEMPORAL LANDSAT-8 IMAGERIES

Hits: 80
Research areas:
Year:
2017
Type of Publication:
Article
Keywords:
Spectral consistency, relative normalization, pseudoinvariant features, constrained regression
Authors:
Muhammad Aldila Syariz, Bo-Yi Lin Chao-Hung Lin
Abstract:
Radiometric normalization is necessary to do since the acquired satellite images contain errors due to several factors such as atmospheric effect. For most historical experiments, the associated atmospheric properties may be difficult to obtain even for planned acquisitions. A relative normalization method is an alternative whenever absolute radiances are not required. Key to the relative normalization is the selection of pseudo-invariant features (PIFs), a group of pixels which is statistically nearly-constant over two images in two different acquisition dates.Several methods; e.g. manual selection, histogram matching, and principal component analysis; had been proposed for extracting PIFs. Yet, a spectral inconsistency, a change in pixel’s spectral signature before and after normalization, is detected wheneverthose PIFs extraction methods, associated with a regression process, are performed. To overcome these shortcoming, the commonly used PIFs selection, called multivariate alteration detection (MAD), is utilized as it considers the relationship among bands. Further, a constrained regression is adopted to enforce the normalized pixel’s spectral signature to be consistent as possible. These approach is applied to multi-temporal Landsat-8 imageries. Moreover, spectral distance and similarities are utilized for evaluating the consistency of the normalized pixel’s spectral signature.
You are here: Home ACRS ACRS Overview Proceedings spectrally-consistent-relative-radiometric-normalization-for-multitemporal-landsat-8-imageries