Robust super-resolution depth imaging via a multi-feature fusion deep network

The number of applications that use depth imaging is increasing rapidly, e.g. self-driving autonomous vehicles and auto-focus assist on smartphone cameras. Light detection ranging (LIDAR) via single-photon sensitive detector (SPAD) arrays an emerging technology enables the acquisition images at high frame rates. However, spatial resolution this typically low in comparison to intensity recorded by conventional To increase native from a SPAD camera, we develop deep network built take advantage multiple features can be extracted camera’s histogram data. designed for camera operating dual-mode such it captures alternate rates, thus system does not require any additional sensor provide images. then uses down-sampled histograms guide up-sampling depth. Our provides significant image enhancement denoising across wide range signal-to-noise ratios photon levels. Additionally, show applied other data types data, demonstrating generality algorithm.