In this paper we present a set of multispectral images covering the visible and near-infrared spectral range (400 nm to 1050 nm). This dataset intends to provide spectral reflectance images containing daily life objects, usable for silicon image sensor simulations. All images were taken with our acquisition bench and a particular attention was brought to processings in order to provide calibrated reflectance data. ReDFISh (Reflectance Dataset For Image sensor Simulation) is available at: http://dx.doi.org/10.18709/perscido.2020.01.ds289.
The research problem is to find an effective enhancement method for enhancing raw underwater color images. In underwater, as depth increases the high wavelength regions of the light spectrum are absorbed by the water and the light spectrum consists only of low wavelength regions such as green and blue and therefore, the image captured underwater looks green or greenish blue. This paper proposes an enhancement algorithm for improving the quality of raw underwater images by the method of alpha-rooting by two-side 2-D quaternion discrete Fourier transform (QDFT) with color correction done by multiscale retinex (MSR). The results of proposed enhancement are compared with the alpha-rooting method, by transforming color images to 2-D grayscale images. The enhancement are measured with reference to the metric color enhancement measure estimation.