Non-line-of-sight imaging with absorption backprojection

Abstract

The backprojection algorithm is a commonly used technique in non-line-of-sight imaging, but it is plagued by artifacts that significantly reduce the quality of the final reconstruction. To address this issue, we refer to the absorption of surrounding dust by celestial bodies and propose the absorption backprojection (ABP) algorithm by imitating the law of gravitation. This algorithm absorbs low confidence value voxels in hidden space into high confidence value voxels so that the majority of the artifacts surrounding the object are eliminated. The results of our study demonstrate that the ABP algorithm is highly effective in reducing the artifacts of backprojection in both simulation and laboratory experiments. Furthermore, it achieves better reconstruction results than other backprojection reconstruction algorithms, including filtered backprojection and error backprojection.