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

De-hazing process with TC4 in IMAGINE

At first, we implemented a robust image-based regression process (which could be considered also a result of its own). It can be used to give the linear regression between 2 images or image bands.

In de-hazing or relative matching, one possibility is to calculate a plane-like correction, or, even better, a haziness index image. Here we need Tasseled Cap transformation’s (TC) fourth component (Crist et al 1986, Lauvreau 1991). It is defined separately for Landsat 4 & 5 gray tones. Our de-hazing process can be used also for Landsat 7 data, if DN values are first transformed to correspond Landsat 5 DNs.

A robust regression model was implemented into ERDAS IMAGINE environment. The whole image area can be used if wanted so. This is a remarkable result, since IMAGINE has been lacking such a robust process. Often there is some overflow effects. Our process also considers the background value 0 and saturated values 255 in a consistent way in all phases of the model.

The de-hazing process uses the regression and eliminates the haze effect in Landsat satellite images. In the discussion lists, ERDAS or Leica Photogrammetry Suite IMAGINE users have asked for such methods. We have used the process for some change detection and mosaicking projects.

The TC4 dehazing process is implemented as a 2-phase process now in IMAGINE Graphical Modeler. It is possible to further develop a graphical interface (and a smooth multi-band process, as with C-factor).

Reported by

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