Bitstream-Based Perceptual Quality Assessment of Compressed 3D Point Clouds.

Area of Science:

• Computer Vision
• Signal Processing
• Multimedia Engineering

Background:

• Increasing demand for compressed 3D point cloud streaming under bandwidth constraints.
• Need for accurate and efficient quality assessment for optimizing end-user Quality of Experience (QoE).
• Limitations of existing methods requiring full or partial decoding of compressed data.

Purpose of the Study:

• To develop a bitstream-based no-reference (NR) model for perceptual quality assessment of 3D point clouds.
• To enable quality evaluation without full decoding of the compressed data stream.
• To provide an efficient method for assessing 3D point cloud quality.

Main Methods:

• Established a relationship between texture complexity, bitrate, and texture quantization parameters using an empirical rate-distortion model.
• Constructed a texture distortion assessment model based on texture complexity and quantization parameters.
• Combined texture and geometric distortion models (derived from Trisoup encoding parameters) to create the streamPCQ model.

Main Results:

• The streamPCQ model accurately predicts perceptual quality of compressed 3D point clouds.
• Demonstrated highly competitive performance compared to existing Full-Reference (FR) and Reduced-Reference (RR) methods.
• Achieved significant computational cost savings compared to traditional quality assessment methods.

Conclusions:

• The proposed streamPCQ model offers an efficient and accurate solution for NR 3D point cloud quality assessment.
• Bitstream-based analysis enables faster and more resource-friendly quality evaluation.
• The model is vital for optimizing QoE in bandwidth-constrained 3D streaming applications.