Multispectral imaging captures images illuminated at wavelengths from ultraviolet to infrared to recover hidden features. Images acquired at different times or of a moving object must be registered. This is challenging due to inconsistent lighting and the sporadic appearance of features across wavelengths. Therefore, feature-based and intensity-based methods often fail. Three registration techniques based on the mean squared differences measure, mutual information and phase-correlation were evaluated and compared for registering affine distortions in multispectral images.
This is poster number 60 in Poster Session