AOTF based molecular hyperspectral imaging system and its image pre-processing method

Abstract

Utilizing both spectral and spatial information, hyperspectral images provide more detail than the single point spectroscopy or traditional images. With the improvement of hyperspectral imaging instrumentation, hyperspectral data analysis techniques are getting more attention. However, desired analysis and results depend strongly on the quality of image preprocessing. In this paper, we obtained hyperspectral images through an Acousto-Optic Tunable Filter (AOTF) based molecular hyperspectral imaging (MHI) system. Then the Lambert-Beer law based pre-processing method was presented to calibrate molecular hyperspectral images. This method combines the advantages of spectral correction and spatial denoising. We applied K-Means classification algorithm and IsoData classification algorithm to the preprocessed blood cell images and the experimental results prove the performance of the preprocessing method which is useful for the further classification.