Method for coding and an apparatus

What is this patent about?

’833 is related to the field of video compression, specifically the efficient encoding and decoding of motion information. In modern video codecs, motion vectors are often predicted from neighboring blocks to reduce the amount of data that needs to be transmitted. A common technique is the use of a merge list, which contains a set of motion vector candidates from which the encoder selects the best match, signaling only the index of that candidate to the decoder.

The underlying idea behind ’833 is to optimize the construction of the merge list by performing a limited redundancy check that is spatially aware. Instead of comparing every new candidate against every existing candidate in the list—which is computationally expensive—the system uses the geometric location of the neighboring blocks to determine a specific subset of prior candidates for comparison. This targeted approach identifies and removes redundant motion information without the overhead of exhaustive pair-wise matching.

The claims of ’833 focus on a method for generating a merge list by selecting a potential spatial motion vector candidate and determining a location-based subset of existing candidates for comparison. The process involves performing an equivalence check or similarity metric between the potential candidate and only those in the determined subset. If a match is found, the candidate is excluded from the list, ensuring that the final signaled index refers to a unique set of motion parameters.

In practice, the invention works by evaluating neighbors in a specific order, such as left, top, and corner blocks. For example, a candidate from a block on the left might only be compared to the block above it if certain splitting conditions are met. By using the spatial relationship between blocks to prune the candidate list, the encoder and decoder can maintain identical lists with fewer operations, which is critical for real-time hardware implementations where memory bandwidth and logic gates are limited.

This approach differs from prior solutions that either performed full redundancy checks or used fixed, non-adaptive candidate sets. By integrating the block partitioning mode (such as vertical or horizontal splits) into the pruning logic, the invention prevents the inclusion of redundant motion vectors that are mathematically likely to be identical due to the way images are subdivided. This results in a more compact merge list that improves compression efficiency while reducing the computational burden on the decoder.