Depth Map Coding Using Predetermined Patterns

The patent segments the smooth area coding process into distinct coding patterns (horizontal, vertical, diagonal) that can be independently selected and applied. Each pattern defines specific neighboring sample positions, breaking down the complex candidate list construction into simpler, pattern-based operations that reduce processing complexity while maintaining coding efficiency.

The patent performs preliminary action by pre-defining multiple coding patterns with specific neighboring sample positions before the actual reconstruction process. These patterns are established in advance based on the characteristics of smooth areas, allowing the decoder to simply select and apply the appropriate pattern without performing complex real-time analysis, thus maintaining accuracy while reducing processing complexity.

Instead of constructing a candidate list and searching for the best candidate (traditional approach), the patent inverts the approach by pre-defining coding patterns that directly specify which neighboring samples to use. This inversion eliminates the need for complex candidate list management and searching, reducing device complexity while maintaining versatility through multiple predefined patterns that can adapt to different smooth area orientations.