Adaptive Reference Frame Padding for Video Coding Memory Efficiency

The frame padding area size is made dynamic and adaptive based on reference picture characteristics. The system determines different padding sizes for different reference pictures using formulas that consider picture width, height, and type (e.g., 1/16th of width for inter-coded pictures, 1/8th for intra-coded pictures). This dynamic adaptation resolves the contradiction by optimizing memory usage per reference picture while maintaining or improving coding performance through appropriate padding for each case.

The invention changes the parameter of frame padding area size from a fixed constant to a variable determined by reference picture attributes. By calculating padding size as a function of picture dimensions and coding type, the system achieves better performance-memory tradeoff. Smaller pictures receive proportionally smaller padding areas, reducing total memory requirements while ensuring adequate padding for accurate motion vector prediction.

The padding area size parameter is changed from fixed to adaptive based on reference picture characteristics. By using formulas that calculate padding as a proportion of picture dimensions (e.g., width/16, height/8), the system achieves motion vector accuracy proportional to picture size without uniformly increasing processing overhead for all pictures. This resolves the contradiction by making precision improvement scalable and efficient.

Different padding area sizes are applied locally to different reference pictures based on their specific characteristics (width, height, coding type). Instead of applying uniform padding to all references, the system tailors padding size to each reference picture's needs, achieving necessary motion vector accuracy for each case while minimizing total processing overhead and memory usage across the entire video sequence.