Video Coding Region Flipping for Prediction Accuracy

The frame is divided into multiple regions (e.g., left region, right region, top region, bottom region) instead of processing as a single block. Each region can be processed independently with its own processing order, allowing reference samples to be available within each region while maintaining overall coding efficiency.

The processing order is made adaptive rather than fixed. The system can dynamically select different processing orders (e.g., left-to-right, right-to-left, top-to-bottom, bottom-to-top) based on the specific coding requirements and reference sample availability, optimizing coding efficiency for each scenario.

By segmenting the frame into multiple regions, the patent ensures that reference samples are available within each region for angular prediction modes. The processing order within each region can be optimized to guarantee reference sample availability, thereby improving reliability while maintaining prediction mode diversity.

Different regions can have different processing orders optimized for their specific characteristics. This allows each region to have reference samples available in the optimal position for angular prediction, improving overall reliability while maintaining the versatility of different prediction modes across the frame.

The patent exploits symmetry axes of regions to flip them, transforming asymmetric regions into more symmetric configurations that are easier to process and code. This reduces discontinuities at region boundaries and improves coding efficiency by leveraging the symmetry property.

Instead of processing regions in the conventional left-to-right, top-to-bottom order, the patent can flip regions horizontally or vertically (e.g., right-to-left, bottom-to-top) to optimize the processing order. This inversion strategy reduces discontinuities and improves coding efficiency by aligning reference samples more effectively.