Multiple band short wave infrared mosaic array filter

Description (excerpt)

Multiple Band Short Wave Infrared Mosaic Array Filter CROSS-REFERENCE TO RELATED APPLICATIONS [0001] This application claims priority to U.S. Provisional Patent Application Serial No. 62/050,117, entitled "MULTI-BAND SHORT-WAVE INFRARED MOSAIC ARRAY FILTER," to Yetzbacher al., which was filed on 13 September 2014 and is incorporated herein by reference. FIELD OF THE INVENTION [0002] The present invention relates to multispectral image collection. More particularly, the present invention involves multispectral image collection in short-wave infrared spectral range. BACKGROUND OF THE INVENTION [0003] Multi-spectral and hyper-spectral imagery has been used to detect and identify vegetation, minerals, chemicals, and types and health of biological tissues. To date, the majority of hyperspectral and multispectral imaging systems have employed line-scanning push-broom instrument designs, using an Offher imaging spectrometer design. Some designs use a specially modified stabilized gimbal design to allow a line scanning spectrometer to collect hyperspectral data at highly oblique angles, such as the MX-20SW hyperspectral imager developed by the Naval Research Laboratory and built by Brandywine Photonics LLC). However, the line-scanning nature of these instruments limits the revisit rate over a designated area due to the time required to build the full hyperspectral data cube. Desires for persistent hyperspectral data and full motion video hyperspectral data have led to alternative designs, such as PHIRST Light, that use a liquid crystal tunable filter to persistently image an entire scene at near video rates while scanning the liquid crystal filter through the desired spectra). The disadvantage of this approach is that the full spectra from the scene are not collected simultaneously in time. The computed-tomography imaging spectrometer ("CTIS") allows for single shot collection of both spatial and spectral information at the expense of optical and computational complexity, and focal-plane array real-estate. A more recent approach has been demonstrated using plenoptic principles with relatively simpler optical elements. This specialized plenoptic camera utilized spectral elements on the objective lens and a lenslet array on the detector to trade spatial resolution for spectral resolution. However, the device requires a lens specifically matched to the lenslet array and focal-plane array design, and additional computation is required to account for lens-specific aberrations. Lastly, conventional color filter array ("CFA") technology places a repeating, or mosaic, pattern of filter elements above blocks of detector pixels on a large focal-plane array. Image processing is then performed on this mosaic pattern image to generate a complete hyperspectral data set for multiple locations within the imaged area. The CFA design removes the restrictive lens tailoring that is found in all line scanning and lenslet systems. This allows for greater ability to change lenses in the field and to quickly adapt cameras to a variety of imaging needs. [0004] Color filter arrays are commonly used to sense color information using two- dimensional and linear photosensitive arrays. See, e.g., U.S. Patent No 3,971,065 to Bayer, which is incorporate herein by reference. Most commercial color visible camera systems use a Bayer color filter array consisting of a repeating 2x2 pattern of individual red, blue, and green filtered pixels, although many variants of the 3-color visible color filter array are known. See, e.g., U.S. Patent Application Publication No. 20070024931 to Compton et al. and U.S. Patent Application Publication No. 20070145273 to Chang, which are both incorporate herein by reference. Visible- band color filter arrays typically use absorptive materials for their filter elements, although there are examples using dielectric thin-film filter elements. See, e.g., U.S. Patent No 7,648,808 to Buchsbaum. [0005] Related art exists to extend color filter arrays to more bands than found in traditional 3-color Bayer filter arrays. For example, based on recovering high-spatial frequency luminance and low spatial-frequency chrominance, Hirakawa and Wolfe suggested arbitrary numbers of spectral bands could be achieved by mosaic array filters that were linear combinations of the Bayer pattern. See, e.g., Hirakawa, K. et al., "Spatio-Spectral Color Filter Array Design for Optimal Image Recovery," Instutute of Electrical and Electronics Engineers (IEEE) Transactions on Image Processing, Vol. 17, No. 10, pp. 1876-1890 (2008), incorporated herein by reference. However, in the short-wave infrared wavelength range, correlations between color channels are not guaranteed. Further, chrominance and luminance are not well defined in the SWIR, being derived from human vision characteristics. Miao et. al. have given an alternative way to design multi- spectral CFAs that is independent of human vision; they base CFA design on the probability of appearance of a band. See e.g., Miao, L., et. al., "Binary tree-based generic demosaicking algorithm for multispectral filter arrays," IEEE Transactions on Image Processing, Vol. 15, No. 1 1, p. 3550 (2006), incorporated herein by reference. Miao et al.'s design scheme is, however