The paper shows that the moment invariants proposed recently in this journal by Hjouji et al. (J Math Imaging Vis 62:606–624, 2020) are incomplete, which leads to a limited discriminability. We prove this by means of circular projection of the image. In a broader context, we demonstrate that completeness of the invariants leads to a better recognition power.

中文翻译：

---

完备集和不完备集

该论文表明，Hjouji 等人最近在该期刊中提出的矩不变量。(J Math Imaging Vis 62:606–624, 2020) 是不完整的，这导致了有限的可辨别性。我们通过图像的圆形投影来证明这一点。在更广泛的背景下，我们证明了不变量的完整性会带来更好的识别能力。