Comparison of quality criteria for hyperspectral images


These results were presented during SPIE Electronic Imaging, January 20th, 2005, in San Jose (California). The proceeding Comparison and evaluation of quality criteria for hyperspectral imagery is available on-line or pdf version.

For further information, you can contact me at emmanuel.christophe@tesa.prd.fr.

Signification for symbols on the curves is available here.

moffett2 for SAM classification

Aviris hyperspectral image on Moffett Fields presenting uniform areas (salt evaporators). The results of Spectral Angle Mapper classification to different degradations is compared with several quality criteria (distortion measurements).

Full results (about 400 kB)

moffett3 for SAM classification

Aviris hyperspectral image on Moffett Fields presenting urbans areas (roads, buildings, vegetation). The results of Spectral Angle Mapper classification to different degradations is compared with several quality criteria (distortion measurements).

Full results (about 400 kB)

moffett2 for Mahalanobis classification

Aviris hyperspectral image on Moffett Fields presenting uniform areas (salt evaporators). The results of Mahalanobis classification to different degradations is compared with several quality criteria (distortion measurements).

Full results (about 400 kB)

moffett2 for Maximum Likelihood classification

Aviris hyperspectral image on Moffett Fields presenting uniform areas (salt evaporators). The results of Maximum Likelihood classification to different degradations is compared with several quality criteria (distortion measurements).

Full results (about 400 kB)

Railroad Valley for PCA

Aviris hyperspectral image on Railroad Valley presenting mineral areas. The results of Principal Component Analysis (number of band to keep 99.9% of the energy) to different degradations is compared with several quality criteria (distortion measurements).

Full results (about 400 kB)

Railroad Valley for Anomaly detection

Aviris hyperspectral image on Railroad Valley presenting mineral areas. The results of Anomaly detection to different degradations is compared with several quality criteria (distortion measurements). These results are not stable enough to draw any conclusion

Full results (about 400 kB)

Back

11-2004
- E. Christophe -