Abstract

Single-pixel imaging (SP) uses coded aperture (CA) elements to capture multiple spatially modulated versions of a scene in a single detector. In real SP implementations, spatial light modulator (SLM) devices are employed to generate the CA modulations. To obtain noiseless discrete images from the SP measurements, it is required to oversample the number of pixels in the objective image. Alternatively, by exploiting compressive sensing theory and designed CA patterns, the SLM-based SP system reduces the projections needed for quality reconstructions, where the CA patterns are established with the aim of emulating orthonormal bases that minimize the correlation between the shots. Nevertheless, in practice, the frame rate of the SP system is restricted by the SLM device, which limits real-time applications. To overcome this, the SLM-based patterns are replaced by etching CA structures over a circularly shifted mask (S-CA), which are introduced in the SP optical path for higher frame rate acquisition. Yet these S-CA patterns present correlated shift modulations and, in turn, yield inaccurate reconstructions. This work introduces an iterative strategy for designing the spatial S-CA pattern structure in SP imaging. The proposed method determines the spatial entries of the S-CA by minimizing the correlation between the pattern shifts, in which the quantity and step size of the S-CA displacements are restricted. Numerical simulations using the proposed design demonstrate an improvement in terms of peak signal-to-noise ratio (PSNR) of up to 2.5 dB, when compared to non-designed CA-S structures, in a compression ratio of 0.25. © 2024 SPIE.