Global Gradient Estimation of Hyperspectral Images for Registration Refinement in Multimodal Microspectroscopy

This contribution introduces a new methodology for the automated estimation of grayscale representations for hyperspectral images (HSI) in the context of multimodal vibrational microspectroscopic imagery. The purpose of the estimated image is to enable a refinement in intensity-based registration of already coarsely registered HSI. The proposed approach derives and fuses gradient information that are globally distributed in the spectral domain using image data from conventional brightfield microscopy (BFM) as a guidance and anchor image for indirect refinement of HSI registration. It is demonstrated that the global gradient image estimated by solving two different optimization problems, reliably improves device-based registration of HSI generated by Raman microspectroscopy (RMS) and infrared microspectroscopy (IRMS).