Wspólnotowy Serwis Informacyjny Badan i Rozwoju - CORDIS


OMEGA Streszczenie raportu

Project ID: EVK2-CT-2000-00069
Źródło dofinansowania: FP5-EESD
Kraj: Finland

Iterative mean & standard deviation matching in IMAGINE

Often in change detection, relative matching of digital images is done using regression methods. But for high resolution images, there may be geometric differences due to different viewing geometry, different shadowing, DEM errors, BDRF effects, etc., which cause the rectified images not to be optimal for regression.

We implemented the mean & standard deviation matching. The method corresponds to distribution (histogram) matching, according to 1st and 2nd moment, or brightness & contrast matching.

Compared to regression, it may be better in certain cases because the above reasons and due to added features:
- We detect and exclude the areas that have changed (i.e. that after the transformation still have big enough brightness difference, not to be considered normal).

- We made the process iterative, so that after defining the linear image-to-image coefficients we can re-threshold the changed areas and re-calculate the coefficients, until they remain stable.

In IMAGINE environment (GUI), the user can select the images, the area used for the statistics, the initialisation method, plus other parameters like the change threshold, exaggeration coefficient for the update vector of the coefficients.

The difference image can also be produced here, by just pressing the button.

Reported by

WM-data Novo
Valimotie 15 B
00381 Helsinki
See on map
Śledź nas na: RSS Facebook Twitter YouTube Zarządzany przez Urząd Publikacji UE W górę