Merge Flag Coding With Non-Empty Candidate Lists in Video Decoding

The patent segments the merge flag encoding process into two independent parts: (1) always encoding the merge flag for every inter-predicted PU regardless of candidate list size, and (2) separately handling the candidate list construction. This segmentation eliminates the coupling that causes error propagation, allowing the decoder to always expect a merge flag and independently construct the candidate list, thereby improving error resilience while maintaining coding efficiency through alternative signaling methods.

The patent inverts the conventional approach by always encoding the merge flag instead of conditionally encoding it. This inversion simplifies the parsing logic because the decoder no longer needs to check candidate list size to determine whether to expect a merge flag. The coding flexibility previously achieved through conditional encoding is replaced by alternative methods such as always signaling the flag and using different candidate list construction rules, thereby reducing parsing complexity while maintaining adaptability.

The patent applies beforehand cushioning by ensuring the merging candidate list is never empty through pre-defined rules for candidate selection and list construction. This protective measure prevents decoding errors that would occur with empty lists while minimizing overhead by establishing clear, deterministic rules for candidate inclusion. The decoder can always safely access at least one candidate, preventing errors without requiring complex error handling mechanisms that would increase overhead.