Spectral filter array for multispectral imaging

Conventional multispectral imaging (MSI) is achieved by utilizing a spectral filter array--a set of predetermined narrowband spectral filters spatially multiplexed over an array of pixel sensors, which necessitates different filters to be designed for each application. In this dissertation, we propose a fundamentally different approach to multispectral imaging known as the Fourier Multispectral Imaging (Fourier MSI). The proposed method utilizes broadband multichroic filters with sinusoidally varying transmittance as a function of wavenumber. Unlike narrowband measurements, these sinusoidal filter measurements largely avoid aliasing that contaminates the spectra while undersampling with narrowband filters. Because of this, Fourier MSI provides a better recovery from discrete filter measurements and preserves the spectra features over the entire detecting wavelength range. We designed and fabricated these sinusoidal filters in both bulk and pixel formats, built up multispectral imaging system with the manufactured filters and made spectral imaging measurements with numerous targets like multispectral LED array, color checker, etc. The measurements show that spectral features such as reflection and absorption peaks are well preserved with this technique. Compared to multispectral systems based on narrowband filters, the Fourier MSI system generalizes well to applications where we lack a priori knowledge of the expected spectral content, which makes it a versatile technique for a wide range of multispectral imaging applications