After getting RGB images from a hyperspectral image using 5 different techniques, how can one represent the quality of results quantitatively? A quick and handy solution is required.
You have an image with 3 bands combination (RGB) from many channels of a hyperspectral imagery. You did not described which five techniques were applied. In general standard quality tests used in image analysis is peaqk signal to noise ratio (PSNR). If you want to compare relative information gained or lost from the five different images you could also try Kullback-Liebler Distance (KDL).
Do you have a ground truth image available? If yes, you can use peak-signal-to-noise ratio or structural similarity. Otherwise you need no-reference, i.e. blind metrics.
@ Richard
I am checking RGB images obtained from hyperspectral images by using techniques like PCA and CMF stretching, and I want to compare the results. There is no reference image and no ground truth available for quality assessment and that's my question that how to compare those results quantitatively?
Such methods exist, but they have different accuracy and may not reflect the various features of human vision. By the way, check Matlab Central.
- Badges
- Science topic
- Similar topics
- Geoscience
- Cartography
- Mapping