Opt. Express 22, 4985–4994 (2014)

Various multispectral systems have been developed that provide detailed data in several wavelength bands, but for many applications the experimental instrumentation is too complex and the information too detailed to be practical and cost effective. Now, researchers based in the USA and China have developed a simple multispectral camera that can simultaneously acquire four spectral images. It employs a complementary metal–oxide–semiconductor (CMOS) sensor with a four-channel Bayer pattern. As both CCD (charge coupled device) and CMOS photodetectors are sensitive to near-infrared (NIR) radiation, most visible-light systems employ a red, green and blue colour filter array to obtain colour information and an infrared filter to block NIR radiation. In contrast, instead of rejecting NIR light, the new system employs a four-colour filter array (red, green, blue and NIR) to obtain both visible and NIR information at the sensor. The researchers have also developed a colour separation algorithm to extract this information from a single image. They demonstrate three potential applications of this system, namely shadow removal, portrait enhancement (where NIR information is used to obtain a smoother skin appearance) and vein enhancement (useful for vein detection systems). In the future, the team intends to improve the colour correction algorithms by using a look-up table to obtain more accurate colour information.