Scattering suppression in single-photon imaging based on the combined spatial denoising and K-nearest neighbors algorithm.

Imaging through scattering media has gained significant research attention in recent years. LiDAR technology has high imaging resolution, long measurement range, high sampling rate, and rich detection information. However, when scattering media are present in the light path, the loss of laser energy increases, and the issue of scattering noise becomes particularly prominent. To address this challenge, we propose a scanning-based photon-limited imaging system for imaging through scattering media. The system employs a point-to-point scanning strategy and utilizes single-photon detectors to detect extremely weak signals after passing through the scattering media. Since the images captured by the proposed system are always subject to Poisson noise at unknown levels, we further propose an algorithm that combines spatial denoising and K-nearest neighbors (SDKNN) based on single-photon time-resolved technology. This approach extracts ballistic photons from the raw data to enhance imaging performance. We constructed a large-aperture coaxial single-photon LiDAR system and set up a chamber filled with dense fog in the optical path. Experimental results show that the proposed technique can successfully image targets in scattering media with high attenuation coefficients, achieving good results even under conditions with attenuation coefficients as high as 8.38. The algorithm effectively suppresses scattering noise and is suitable for reliable imaging under harsh weather conditions.