Decoder-Side Intra Mode Derivation Using Neighboring Block Templates

The decoder autonomously derives the intra prediction mode by analyzing templates from reconstructed neighboring blocks without requiring explicit mode signaling from the encoder. The decoder performs gradient calculations and mode selection independently, allowing the encoder to simply signal a flag indicating whether TIMD is applied, thereby improving compression performance while minimizing the increase in device complexity.

The intra prediction mode is derived in advance during the decoding process using pre-reconstructed neighboring blocks as templates. By performing the mode derivation before final prediction application and using already-available reconstructed data, the method improves compression efficiency without requiring additional complex real-time calculations during critical decoding paths.

Different templates from different neighboring blocks are used to derive intra prediction modes for different regions. The method selectively applies templates based on local characteristics of neighboring blocks, using gradient analysis to determine which templates provide the best local match. This improves prediction accuracy for each local region while avoiding the need to process all possible templates uniformly, thus reducing overall decoding time.

The method uses a limited set of relevant templates from neighboring blocks rather than exhaustively searching all possible prediction modes. By performing gradient calculations only on selected templates that show promise based on neighboring block characteristics, the method achieves sufficient prediction accuracy without the time cost of exhaustive template evaluation.