Occlusion-Aware Depth Map Coding Optimization Using Allowable Depth Map Distortions.

Area of Science:

• Computer Vision
• Image Processing
• Video Coding

Background:

• Depth map coding is crucial for view synthesis in 3D video.
• Optimizing rate-distortion for occlusion-inducing pixels is challenging due to complex distortion estimation and occlusion order changes.
• Existing methods struggle with accurate synthesis distortion estimation and maintaining occlusion order.

Purpose of the Study:

• To propose an efficient depth map coding approach for occlusion-inducing pixels.
• To minimize overall synthesis distortion in the presence of occlusion.
• To develop a method invariant to occlusion order changes.

Main Methods:

• Derivation of allowable depth level change ranges with theoretical and geometrical proofs.
• Formulation of an optimization problem to select depth distortion for minimizing synthesis distortion.
• Development of a dynamic programming algorithm for optimal depth distortion selection.

Main Results:

• Demonstration of the unicity and occlusion order invariance properties of the allowable depth distortion range.
• Significant performance improvement shown compared to state-of-the-art depth map coding optimization schemes.
• Effective minimization of synthesis distortion for occlusion-inducing pixels.

Conclusions:

• The proposed depth map coding approach efficiently handles occlusion-inducing pixels.
• The method provides theoretical guarantees and demonstrates practical performance gains.
• This technique advances depth map coding for improved 3D view synthesis.