Confocal Non-Line-of-Sight (NLOS) imaging is a technique that allows the reconstruction of hidden objects that are not directly visible to the imaging system. This method typically involves the use of confocal scanning and time-of-flight (ToF) measurements to capture indirect light reflections from hidden scenes.

Unlike traditional methods, circular and confocal NLOS (C2NLOS) employs 1D circular scanning for speed and efficiency but faces limitations due to spatial bias, which affects the quality of the reconstruction results. We propose E-C2NLOS (Enhanced C2NLOS), a framework in which we directly enhance the measured information. Our framework includes Cropping and Stitching modules for effective image processing, and our method outperforms the baseline methods.

The reconstruction results achieved through our framework exhibit higher quality, particularly in effectively reducing artifacts in the background. And our framework requires only about 3% to 7% of the GPU memory demanded by the original method, significantly reducing hardware requirements and enhancing practicality.