Ultrafast single-pixel optical imaging based on multimode interference and compressed sensing (Conference Presentation)

Abstract

Single-pixel imaging based on structured illumination and compressed sensing has opened a new way to compress massive imaging data volume and significantly reduce the cost of image sensor without sacrificing imaging quality. However, conventional structured illumination methods based on digital micro-mirror device (DMD) or a liquid crystal based spatial light modulator (SLM) fall short in fresh rate, making it a real challenge for high-speed imaging applications, which are however of paramount importance in studying dynamic phenomena in living cells, neural activity, and microfluidics, and capturing important rare events. In this work, we propose and demonstrate a new approach for ultrafast (20 Mfps) structured illumination single-pixel imaging using light beam speckles out of a multimode fiber due to multimode interference. Our experimental results show that the excited high-order modes, and hence the multimode interference, are strongly wavelength-dependent. Update of the random speckle patterns can be easily obtained by sweeping the incident wavelength. Ultrafast wavelength sweeping is achieved by stretching ultrafast optical pulses from a mode-locked laser using chromatic dispersion. Extremely broad bandwidth and small wavelength step guarantee a good number of illumination patterns. By measuring multiple dot products of a sparse image with a set of known speckle based random, the image can be reconstructed using an L1 minimization algorithm. The most significance of this completely new design is that multiple (up to thousands) structured illumination measurements can be carried out within a single pulse period, enabling ultrafast pulse-by-pulse imaging. Moreover, thanks to structured illumination and compressed sensing, the proposed structured illumination single-pixel imaging system offers much higher imaging resolution than existing ultrafast photonic time stretch imaging systems for the same captured data size.