Compressive imaging beyond the sensor's physical resolution via coded exposure combined with time-delay integration

Abstract

Compressive imaging has been widely involved in high-dimensional data acquisition such as hyperspectral, ultrafast, and depth imaging. As for its application in high-resolution imaging beyond the sensor's physical resolution, likewise, sub-pixel information needs to be spatially modulated by the essential encoding devices before being projected onto the sensor, thereby introducing additional devices into the optical path with calibration requirements. Toward this end, we propose a high-resolution compressive imaging framework that implements spatial encoding by combining coded exposure with time-delay integration, exploiting its sub-pixel intra-line charge transfer characteristics. In this way, no additional encoding device except a fluttering shutter is required, thus simplifying the system composition and calibration, with the conventional imaging capability preserved. The dual-arm architecture designed to capture sub-pixel information in two orthogonal directions separately is integrated into a joint forward model, in which the point spread function is also introduced as an efficient part of coding to improve the degraded reconstruction caused by optical aberrations. Based on the mutual coherence criterion, the exposure sequence is optimally designed by the genetic algorithm, achieving a substantial improvement in image quality. The calibration methods for this system are also detailed. Then, the imaging performance affected by both internal configurations and external errors in practical applications are quantitatively analyzed through numerical simulations, as well as the generalizability to different types of images. Finally, we built a proof-of-principle prototype to demonstrate that the proposed framework is a viable and effective alternative for high-resolution compressive imaging.