Color Enhancement for V-PCC Compressed Point Cloud via 2D Attribute Map Optimization
Jingwei Bao, Yu Liu, Zeliang Li, Shuyuan Zhu, Siu-Kei Au Yeung
TL;DR
The paper tackles color fidelity degradation in V-PCC compressed point clouds by optimizing 2D projection attribute maps with a lightweight LDC-Unet and a two-stage transfer-learning workflow that leverages abundant 2D portrait data before fine-tuning on point-cloud projection maps. The approach formalizes the pipeline as $y' = M(y) = M(C(P(x)))$ with back-projection $x' = P^{-1}(y')$, enabling 3D reconstruction improvements. Key contributions include the LDC-Unet architecture with depthwise separable convolutions and residual contextual blocks to reduce parameters, a custom 2D portrait dataset for pretraining, and a transfer-learning strategy that enhances generalization to real point-cloud projection maps. Empirical results on 8iVSLF sequences show PSNR gains for both 2D attribute maps and 3D reconstructions, especially at higher quantization parameters, while achieving competitive performance with fewer parameters than DRUNet. The work improves practical color quality in V-PCC pipelines and points to future extensions for joint optimization of geometry maps under lossy conditions.
Abstract
Video-based point cloud compression (V-PCC) converts the dynamic point cloud data into video sequences using traditional video codecs for efficient encoding. However, this lossy compression scheme introduces artifacts that degrade the color attributes of the data. This paper introduces a framework designed to enhance the color quality in the V-PCC compressed point clouds. We propose the lightweight de-compression Unet (LDC-Unet), a 2D neural network, to optimize the projection maps generated during V-PCC encoding. The optimized 2D maps will then be back-projected to the 3D space to enhance the corresponding point cloud attributes. Additionally, we introduce a transfer learning strategy and develop a customized natural image dataset for the initial training. The model was then fine-tuned using the projection maps of the compressed point clouds. The whole strategy effectively addresses the scarcity of point cloud training data. Our experiments, conducted on the public 8i voxelized full bodies long sequences (8iVSLF) dataset, demonstrate the effectiveness of our proposed method in improving the color quality.