SPLINE-Net: Sparse Photometric Stereo through Lighting Interpolation and Normal Estimation Networks
This addresses the challenge of reducing the number of required images for photometric stereo, which is important for applications in computer vision and graphics, though it appears incremental as it builds on existing network-based approaches.
The paper tackles the problem of sparse photometric stereo for general BRDFs by proposing SPLINE-Net, which uses a lighting interpolation network and a normal estimation network with symmetric and asymmetric loss functions, achieving better performance than existing methods while using only ten images instead of nearly one hundred.
This paper solves the Sparse Photometric stereo through Lighting Interpolation and Normal Estimation using a generative Network (SPLINE-Net). SPLINE-Net contains a lighting interpolation network to generate dense lighting observations given a sparse set of lights as inputs followed by a normal estimation network to estimate surface normals. Both networks are jointly constrained by the proposed symmetric and asymmetric loss functions to enforce isotropic constrain and perform outlier rejection of global illumination effects. SPLINE-Net is verified to outperform existing methods for photometric stereo of general BRDFs by using only ten images of different lights instead of using nearly one hundred images.