Figure 1. Illumination’s difference between conventional single pixel imaging (SPI) and the proposed multispectral single pixel imaging (MSPI).

Due to the response speed gap between a bucket detector (MHz or GHz) and a spatial light modulator (no higher than KHz), the detector can collect a dense sequence of measurements during the elapse of each spatially modulated illumination pattern. In conventional SPI, given a spatial pattern, its light intensity and corresponding measurements are constant. Thus no spectral information can be extracted from the measurement sequence. Differently, both the illumination’s intensity and the measurements are time-varying in MSPI, because the light’s intensity of each spectral component changes sinusoidally over time with different periods. Thus we multiplex the target scene’s spectral information into the measurement sequence during the elapse of each spatial pattern, and can use Fourier decomposition to demultiplex these spectral information. This is the main difference between conventional SPI and the proposed MSPI.