Enhancing Image Reconstruction under Low-Light and Long-Distance Conditions Using Maximum Likelihood Estimation

Single-photon sensors are crucial for image reconstruction, particularly under low light and close proximity, due to their high sensitivity to photon arrival and straightforward data processing. This paper introduces the use of Maximum Likelihood Estimation (MLE) to enhance image reconstruction by optimizing the processing of incomplete pixel data carried by photons. MLE is employed to construct a macroscopic probability model through Mixed Distribution, integrating and iterating over defective data within an image. Additionally, a Loss Function is introduced to train and evaluate this model, enhancing the robustness and accuracy of the reconstruction by extending the iteration time and integrating defective data repetitively. An interval checking mechanism is developed to ensure the accuracy of the reconstructed images, providing a standard to measure the reliability of the computed values and associated variables that correlate positively with the outcome. The application of these methods significantly enhances image parameters such as Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM). Results indicate that MLE is highly effective in identifying the most probable segments of incomplete data, thereby enabling precise image reconstruction in challenging conditions.