BROWSE
Detail
Journal Article
Enhancing Lossless Compression in AI-PCC via Distribution-Aware Feature Extraction
- Authors
- Dowan Kwon, Joon Kwon, Junsik Kim, Kyuheon Kim
- Issue Date
- 2025-12
- Citation
- 방송공학회 논문지, v.30, no.7, pp.1143-1154
- ISSN
- 1226-7953
- Publisher
- 한국방송∙미디어공학회
- Language
- English
- Type
- Journal Article
- Abstract
- The Moving Picture Experts Group (MPEG) is currently standardizing AI-based Point Cloud Compression (AI-PCC). Since AI-PCC reconstructs point clouds using geometric features extracted from the input, the quality of reconstruction improves as these extracted features become more precise. However, although point clouds exhibit variations in their geometric distributions, the current AI-PCC framework extracts features using identical cubic receptive fields. To address this limitation, this paper proposes a distribution-aware feature extraction method that uses an orthotropic rect-kernel extractor/aggregator for dense point clouds to capture linear and planar characteristics, and a sparse dilated convolution extractor for sparse point clouds to reflect point dispersion. The proposed method enhances lossless compression performance in AI-PCC by performing adaptive feature extraction to the structural heterogeneity of point clouds.
- Keyword
- 3D Point Cloud, Geometry Coding, AI-PCC, Feature Extraction, Lossless Coding
- KSP Keywords
- 3D Point Cloud, Adaptive feature extraction, Compression performance, Dilated Convolution, Geometric features, Moving picture experts group(MPEG), Point cloud compression, Quality of Reconstruction, Receptive field, Sparse point, Structural heterogeneity
This work is distributed under the term of Creative Commons License (CCL)
(CC BY ND)
(CC BY ND)
