Distributed successive refinement of multiview images using broadcast advantage

Area of Science:

• Computer Vision
• Image Processing
• Information Theory

Background:

• Environmental monitoring benefits from robust multi-camera systems.
• Successive refinement image coding enables energy savings by allowing progressive quality requests.
• Exploiting correlations in multiview images is key for efficient coding.

Purpose of the Study:

• To propose a novel two-encoder successive refinement scheme for multiview image coding.
• To analyze the scheme's performance against theoretical limits in distributed coding.
• To develop a practical algorithm for stereo-view image coding.

Main Methods:

• Investigated a two-camera setup exploiting broadcast nature and image correlation.
• Proposed a ping-pong-like two-encoder successive refinement scheme.
• Proved theoretical refinability for the bivariate Gaussian case and developed a practical stereo-view algorithm.

Main Results:

• The proposed scheme is proven to be successively refinable on the theoretical rate-distortion limit (Wagner surface).
• A practical algorithm for stereo-view images was developed based on the scheme.
• Simulation results demonstrate performance close to the distributed coding bound.

Conclusions:

• The novel scheme offers efficient and progressively improving image quality for multi-camera systems.
• The approach achieves near-optimal performance within the framework of distributed coding.
• This method enhances energy efficiency and robustness in environmental monitoring applications.