arXiv Computer Vision research summaries for March 13, 2024.

Today's Research Themes (AI-Generated):

• REPAIR tackles the noisy correspondence issue in cross-modal matching by estimating soft labels and replacing mismatched pairs.

• Matching non-identical but semantically similar objects enhanced by a novel weighting scheme of descriptors.

• COT-PCC presents a novel point cloud compression approach using a GAN-based constrained optimal transport formulation.

• The dual-domain regularization model effectively removes ring artifacts in X-ray CT while preserving image quality.

• PNeSM offers a framework for arbitrary 3D scene stylization without scene-specific re-training.