Image Coding With WRP Flags for Geometrically Modified References

The patent applies preliminary action by pre-processing reference pictures through geometric transformation (warped reference picture generation) before motion vector prediction. The WRP Flag information is prepared in advance and stored in candidate blocks, allowing the decoder to efficiently identify and select appropriate candidates without performing complex transformations during the prediction process itself. This pre-computation approach improves compression efficiency while controlling encoding complexity.

The patent introduces WRP Flag information as an intermediary element that mediates between the complex geometric transformation process and the simple motion vector prediction. The flag acts as a marker or indicator that guides the selection of warped reference picture candidates without requiring the decoder to understand or perform the transformation operations. This intermediary mechanism enables efficient candidate selection while maintaining algorithmic simplicity.

The patent applies segmentation by dividing the candidate block structure into distinct components: the original motion information and the WRP Flag indicator. This segmentation allows the system to process and compare only the flag information during candidate selection, rather than performing full geometric transformations on each candidate. The segmented approach enables fast comparison and selection, improving prediction accuracy while minimizing processing time overhead.

The patent uses copying by creating simplified copies of reference picture information with attached WRP Flags, rather than working with the full transformed reference pictures during candidate evaluation. The candidate blocks contain copied motion information and flag indicators that can be quickly compared and selected. Only after selection is the actual geometric transformation applied to the final chosen candidate, significantly reducing processing time while maintaining prediction accuracy.