Exploiting Color Filter Array Sensor NIR Spectral Sensitivity toImprove Color Accuracy: Spectral Reflectance Analysis

摘要

A visible and near infrared (NIR) spectral reflectancerndatabase of commonly imaged objects was developed havingrnminimal spectral redundancy. A spectral analysis wasrnperformed on this database to investigate the relationshiprnbetween NIR and visible spectral reflectance. The correlationrnbetween the digital signals of theoretical NIR channels andrntristimulus values for the database was analyzed to find thernoptimal NIR channel such that this channel could be used tornimprove color accuracy and improve image quality. The closerrnthe NIR channel was located to the visible spectrum, the morernoptimal was the result. A colorimetric image segmentationrnalgorithm was developed to improve this correlation. Thernspectral sensitivities of a typical color filter array (CFA) camerarnwere used as a virtual camera model to test the viability of arnNIR channel and colorimetric segmentation to improve colorrnaccuracy. The simulation results using the database revealedrnthe possibility to exploiting the NIR spectral sensitivities ofrncommon CFA cameras as an additional channel for improvedrncolor accuracy.