The author proposes a novel point cloud compression method, COT-PCC, by framing the task as a constrained optimal transport problem. By incorporating a generative adversarial network and bitrate loss for training, COT-PCC outperforms existing methods in terms of CD and PSNR metrics.
Efficient lossless compression of point cloud geometry and color attributes using sparse tensor-based neural networks.
Leveraging geometric priors enhances human point cloud compression performance.
Pointsoup, an efficient learning-based geometry codec, achieves state-of-the-art compression performance on large-scale point cloud scenes while providing extremely low decoding latency.
A novel hybrid context model, PVContext, that integrates local voxel information and global shape priors from reconstructed point clouds to enable efficient octree-based compression of large-scale point cloud data.
This paper introduces a novel method for compressing large point clouds and performing efficient approximate nearest-neighbor searches using tensor-train (TT) decomposition, a technique commonly used for compressing neural network parameters.
본 논문에서는 대규모 벡터 데이터베이스에서 효율적인 포인트 클라우드 표현과 빠른 근사 최근접 이웃 탐색을 가능하게 하는 새로운 텐서 트레인 기반 포인트 클라우드 압축 방법을 제안합니다.