Servizio Comunitario di Informazione in materia di Ricerca e Sviluppo - CORDIS

Estimating the saturated pixels

Novosat implemented robust image linear regression processes for ERDAS IMAGINE environment. It is used also here in the process for estimating the saturated (255) pixels’ true values, via a regression with the non-saturated Landsat band 4 image values directly. Here Novosat used the robust regression method described in the TC4 haze correction description. Also versions with two explaining bands (for example 4 and 3 when estimating the band 2 or 1) were developed, but not utilizing very many pixels.

The Area-of-Interest (AOI) enables including the small non-saturated gaps and other edge areas of glaciers. Including much of the bright pixels seems to give more reasonable results than with mostly the darker pixels. This is because the bright end of the regression line gets more accurately defined.

The models for estimating the estimated pixels’ true values are based on IMAGINE’s Graphical Modeler processes. They could be commercialised.

Reported by

WM-data Novo
Valimotie 15 B
00381 Helsinki
See on map