A Few Photons Among Many: Unmixing Signal and Noise for Photon-Efficient Active Imaging

Abstract

Conventional LIDAR systems require hundreds or thousands of photon detections to form accurate depth and reflectivity images. Recent photon-efficient computational imaging methods are remarkably effective with only 1.0 to 3.0 detected photons per pixel, but they are not demonstrated at signal-to-background ratio (SBR) below 1.0 because their imaging accuracies degrade significantly in the presence of high background noise. We introduce a new approach to depth and reflectivity estimation that focuses on unmixing contributions from signal and noise sources. At each pixel in an image, shortduration range gates are adaptively determined and applied to remove detections likely to be due to noise. For pixels with too few detections to perform this censoring accurately, we borrow data from neighboring pixels to improve depth estimates, where the neighborhood formation is also adaptive to scene content. Algorithm performance is demonstrated on experimental data at varying levels of noise. Results show improved performance of both reflectivity and depth estimates over state-of-the-art methods, especially at low signal-to-background ratios. In particular, accurate imaging is demonstrated with SBR as low as 0.04. This validation of a photon-efficient, noise-tolerant method demonstrates the viability of rapid, long-range, and low-power LIDAR imaging.

Photon-efficient super-resolution laser radar

“…This is not a low number of photons per pixel comparable to recent works in which deconvolution is not needed, [14][15][16][17] but the results suggest that it is low enough that our explicit Poissonian modeling of observed photon counts improves performance. The flattened photon data cube f (Y) ∈ N nxny×nt is illustrated in image form in Fig.…”

“…[15][16][17] One core assumption behind these existing lowlight depth imaging frameworks is that the spatial spot size of the illumination is small enough that we can assume that each pixel measurement has addressed a different scene patch. However, in practice, this independent-pixel assumption can break down due to non-idealities in the illumination source and imaging conditions.…”

Super-resolution single-photon imaging at 8.2 kilometers

“…According to the calibration of the signal and background noise in the acquired data, the number of signal photons is ∼1 to 6 PPP and the SBR is ∼0.16 to 0.24 (defined within 100-ns gate window similar to ref. [16]). We successfully captured the fine depth maps of the mannequin postures in size of (128 × 128) pixels over 8.2 km in all time.…”

“…We choose a typical scene from the Middlebury dataset [31]. We simulated the results with the detected number of signal photons at 10, 5, 1 PPP, and compared the reconstruction results with state-of-the-art photon-efficient algorithms [15,16,27]. From the results shown in Fig.…”

“…Furthermore, computational imaging algorithms have seen remarkable progress to process the single-photon data efficiently [11]. High-quality 3D structure has been demonstrated in the laboratory environment by an active imager detecting only one photon per pixel (PPP), based on the approaches of first-photon imaging [12], pseudo-array [13,14], single-photon camera [15], unmixing signal/noise [16] and machine learning [17]. These algorithms have the potential to improve the imaging range and quality significantly.…”