Efficient Reconstruction for Dual Profilometry

Fringe projection profilometry has emerged as a widely used non-contact three-dimensional (3D) measurement technique due to its full-field acquisition, high precision, and rapid measurement capabilities. However, its performance degrades significantly when measuring objects with complex surface reflection characteristics, as light undergoes inter-reflection and subsurface scattering during measurement, generating indirect light signals that lead to reconstruction errors. Dual profilometry, a computational 3D profilometry based on dual photography theory, overcomes this limitation by treating each camera pixel as a single-pixel detector to isolate the direct light signals. While theoretically effective, existing implementations need the reconstruction of millions of dual images. To overcome this limitation, we present an efficient reconstruction method that avoids dual image reconstruction. Experimental validation demonstrates that the reconstruction speed can be substantially improved compared to existing dual profilometry implementations, achieved without compromising reconstruction accuracy. The proposed method effectively resolves the efficiency bottleneck that has hindered the practical adoption of dual profilometry, enabling its deployment in practical applications such as industrial inspection and high-resolution biomedical surface characterization.