Separation of Multispectral Raw Image Data for the Analysis of Visible and Near-Infrared Graphic Information

The paper examines the separation of multispectral information from raw image data acquired using a MAPIR Survey3N camera equipped with a cyan / orange / near-infrared filter. Although the camera is primarily designed for agronomic imaging, this study applies it under controlled studio conditions to analyse graphic information. The research is based on the assumption that raw image data contain more original sensor information than JPG files, while also retaining a degree of spectral overlap caused by the Bayer sensor structure and inter-channel crosstalk. For this purpose, a computational workflow was developed to unpack the raw image data, separate them into initial sensor planes, and transform them into three information layers corresponding to spectral peaks at approximately 490 nm, 615 nm, and 808 nm. The workflow was tested on studio samples including marker inscriptions, a plant, and a postage stamp produced using Infra-reDesign technology, which contains both visible and infrared-dependent graphic information. The results indicate that the separated layers provide a clearer basis for comparing the behaviour of colorants and graphic elements than channels extracted from JPG images. The resulting layers should not be interpreted as absolute reflectance measurements, but as relative multispectral representations produced within the same optical-sensor system. The proposed approach demonstrates the potential of compact and affordable multispectral cameras for analysing graphic materials, colorants, and hidden infrared information.