Dimensional Extension Algorithm for Generating Hyperspectral Images Considering Absorption Spectrum Characteristics

A hyperspectral image has wavelength information of several hundreds and can identify and detect objects that are difficult to with RGB images; however, the hyperspectral camera is very expensive. Thus, acquiring hyperspectral images inexpensively will help their widespread use. We propose a method for estimating hyperspectral images from RGB images using a model in which hyperspectral images are represented by a composite of an absorption coefficient matrix and a molar density matrix. In the validity experiment, we verify the difference in the estimation accuracy with the proposed method at four different wavelengths using tomato images.