Perceptual Quality Assessment of Trisoup-Lifting Encoded 3D Point Clouds

Area of Science:

• Computer Vision
• Multimedia Signal Processing
• 3D Graphics

Background:

• No-reference point cloud quality assessment (PCQA) is crucial for real-time monitoring in networks.
• Existing methods often require full decoding, limiting real-time application.
• Trisoup-Lifting is an emerging encoding method for 3D point clouds.

Purpose of the Study:

• Develop the first no-reference (NR) bitstream-layer PCQA model for Trisoup-Lifting encoded point clouds.
• Enable quality assessment without full decoding for real-time network monitoring.
• Establish a dedicated benchmark database for Trisoup-Lifting PCQA.

Main Methods:

• Investigated relationships between texture bitrate per point (TBPP), texture complexity (TC), and texture quantization parameter (TQP) with lossless geometry encoding.
• Estimated TC using TQP and TBPP, then developed a texture distortion model.
• Integrated the texture distortion model with a geometry attenuation factor (based on trisoupNodeSizeLog2, tNSL) to create the streamPCQ-TL model.

Main Results:

• Developed streamPCQ-TL, a comprehensive NR bitstream-layer PCQA model for Trisoup-Lifting.
• Created WPC6.0, the first PCQA database for Trisoup-Lifting, containing 400 distorted point clouds.
• StreamPCQ-TL demonstrated robust performance against existing metrics on multiple datasets, with significantly lower computational cost.

Conclusions:

• The streamPCQ-TL model provides effective and efficient real-time quality assessment for Trisoup-Lifting encoded point clouds.
• The WPC6.0 database serves as a valuable resource for future research in this area.
• The proposed approach advances NR bitstream-layer PCQA for emerging 3D data compression techniques.