[Paper Review] Hybrid Point Cloud Attribute Compression Using Slice-based Layered Structure and Block-based Intra Prediction

Yiting Shao, Qi Zhang+2 more|arXiv (Cornell University)|Apr 28, 2018

Advanced Vision and Imaging18 references3 citations

TL;DR

This paper proposes a hybrid point cloud attribute compression scheme using a slice-based layered structure and block-based intra prediction, integrating adaptive Graph Fourier Transform (GFT) with Lagrangian optimization and multiple reordering scan modes. The method achieves a 29.37% BD-rate gain over the state-of-the-art RAHT system and a 16.37% BD-rate gain over MPEG TMC1 anchor, demonstrating superior rate-distortion performance in intra-frame color attribute coding.

ABSTRACT

Point cloud compression is a key enabler for the emerging applications of immersive visual communication, autonomous driving and smart cities, etc. In this paper, we propose a hybrid point cloud attribute compression scheme built on an original layered data structure. First, a slice-partition scheme and geometry-adaptive k dimensional-tree (kd-tree) method are devised to generate the four-layer structure. Second, we introduce an efficient block-based intra prediction scheme containing a DC prediction mode and several angular modes, in order to exploit the spatial correlation between adjacent points. Third, an adaptive transform scheme based on Graph Fourier Transform (GFT) is Lagrangian optimized to achieve better transform efficiency. The Lagrange multiplier is off-line derived based on the statistics of color attribute coding. Last but not least, multiple reordering scan modes are dedicated to improve coding efficiency for entropy coding. In intra-frame compression of point cloud color attributes, results demonstrate that our method performs better than the state-of-the-art region-adaptive hierarchical transform (RAHT) system, and on average a 29.37$\%$ BD-rate gain is achieved. Comparing with the test model for category 1 (TMC1) anchor's coding results, which were recently published by MPEG-3DG group on 121st meeting, a 16.37$\%$ BD-rate gain is obtained.

Motivation & Objective

• To address the challenge of efficiently compressing irregularly distributed point cloud attributes with high spatial correlation.
• To improve coding efficiency by exploiting spatial redundancy in point cloud attributes through novel structural and predictive tools.
• To optimize transform efficiency using Lagrangian-optimized GFT based on statistical analysis of color attributes.
• To enhance entropy coding performance via adaptive reordering scan modes tailored to residual data distribution.

Proposed method

• A slice-partitioning scheme combined with geometry-adaptive k-d tree partitioning generates a four-layered data structure for hierarchical processing.
• Block-based intra prediction with DC and five angular modes is applied to exploit local spatial correlation; mode selection uses sum of absolute transformed difference (SATD).
• An adaptive transform scheme combines GFT and DCT, with Lagrangian optimization using an off-line derived lambda multiplier based on color attribute statistics.
• The Lagrange multiplier is pre-computed from training data (e.g., Andrew, Phil, Ricardo, Sarah, Queen_frame_0200, etc.) to balance rate and distortion.
• Multiple reordering scan modes are applied to transformed residuals before entropy coding to improve coding efficiency.
• Arithmetic entropy coding is used with uniform quantization, and all coding tools (intra mode, transform mode, scan mode, residual data) are multiplexed into the bitstream.

Experimental results

Research questions

• RQ1 How can a layered data structure be effectively constructed to support hierarchical compression of point cloud attributes?
• RQ2 To what extent can block-based intra prediction reduce redundancy in point cloud attribute data?
• RQ3 Can adaptive GFT-based transform with Lagrangian optimization outperform fixed transform schemes in point cloud attribute compression?
• RQ4 How do multiple reordering scan modes impact entropy coding efficiency for point cloud residuals?
• RQ5 What is the relative contribution of each component (slice partitioning, intra prediction, adaptive transform, scan reordering) to overall coding gain?

Key findings

• The proposed method achieves a 29.37% average BD-rate gain over the state-of-the-art RAHT system in intra-frame point cloud attribute compression.
• On average, the method obtains a 16.37% BD-rate gain compared to the MPEG TMC1 anchor at the 121st meeting.
• The luma component shows a 37.95% BD-rate gain over RAHT, with 26.83% and 23.34% gains for the two chroma components.
• The ablation study confirms that each component—slice partitioning, intra prediction, adaptive transform, and reordering scan—contributes significantly to the overall rate-distortion improvement.
• The method achieves up to 4 dB PSNR gain on the Y component for certain datasets like David and Dimitris compared to RAHT.
• At high bitrates, performance remains competitive, though minor improvements are needed for the transform scheme on some datasets like Longdress.