Analysis of detection bandwidth limitations in time-stretch-based single-pixel imaging systems

Abstract

Single-pixel imaging (SPI) has recently attracted considerable attention in magnetic resonance imaging and remote sensing for its data compression capability and broad operational wavelength range. However, the frame rates of single-pixel cameras are largely limited by the response time of digital micromirror devices (DMDs). To circumvent this challenge, a novel SPI technique called time-stretch-based single-pixel imaging (TSSPI) is proposed, which can increase the imaging speed of single-pixel cameras based on DMD by over 3 orders of magnitude. In the TSSPI system, photodetection bandwidth and pulse repetition rate are critical parameters influencing the quality of reconstructed images. In this paper, we quantitatively analyze their effects on reconstruction accuracy through simulation and experiment for the first time. Photodetectors with various bandwidths are separately used to acquire the same number of measurements for image reconstruction and the peak signal-to-noise ratios of the reconstructed images are calculated to evaluate the reconstruction accuracy. We experimentally demonstrate that the ratio between photodetection bandwidth and pulse repetition rate has a strong impact on reconstruction quality. A threshold for this ratio is estimated and high-quality image reconstruction can be achieved only above this threshold. In our experiment, a 400-MHz detection bandwidth is demonstrated to be an optimized parameter set which can balance the trade-off between the system cost and the reconstruction accuracy with respect to a 50-MHz pulse repetition rate. In addition, we also demonstrate that the ratio is related to the complexity of the illuminated scene.