RDTC optimized compression of image-based scene representations (Part I): modeling and theoretical analysis

Area of Science:

• Computer Vision
• Image Processing
• Multimedia Systems

Background:

• Interactive streaming of image-based scene representations necessitates random access to image data.
• Interframe dependencies in encoding affect transmission, decoding times, and the rate-distortion (RD) tradeoff.
• Classical RD optimization is limited in addressing multiple performance metrics simultaneously.

Purpose of the Study:

• To extend the classical rate-distortion (RD) optimization to a rate-distortion-transmission-complexity (RDTC) tradeoff.
• To develop a theoretical model for the RDTC space, emphasizing decoding complexity.
• To evaluate the impact of client-side caching on RDTC measures.

Main Methods:

• Extended classical RD optimization using hybrid video coding concepts.
• Developed a theoretical model for the RDTC space.
• Incorporated client-side caching effects into the RDTC model.
• Conducted experimental evaluations to validate theoretical predictions.

Main Results:

• Experimental results align with theoretical predictions for the RDTC model.
• Demonstrated that adapting encoding to receiver parameters (computational power, channel throughput) significantly impacts performance.
• Showcased reduced user-perceived delay and storage requirements for RDTC-optimized streams.

Conclusions:

• The proposed RDTC model provides a framework for optimizing image-based scene representations for interactive streaming.
• Adaptive encoding strategies tailored to specific scenarios can yield substantial improvements in perceived quality and efficiency.
• Client-side caching plays a crucial role in the overall RDTC performance of streaming systems.