This is the authors' abstract. We don't add key points for this paper.
Particle shapes of raw, granular materials affect their bulk properties, but 2D measurements can introduce stereological errors. We developed a method to estimate 3D shapes from 2D images without pre-training. First, 3D shapes reflecting measured 2D characteristics are generated using a Fourier shape descriptor. Next, spherical harmonic-based principal component analysis refines them so their 2D projections match observed shapes. Accuracy is fivefold better than using Fourier descriptors alone. This method requires no pre-training, making it applicable to any particles. As a demonstration, we used it to estimate 3D shapes of Martian rocks from Perseverance rover images.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Takao Ueda (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: