Video coding

What is this patent about?

’764 is related to the field of video compression and error resilience, specifically addressing the challenges of temporal predictive coding in unreliable network environments. In standard video coding, such as H.263 or MPEG, pictures are often predicted from previous frames to reduce redundancy. However, if a reference frame is lost during transmission, the decoder may continue to decode subsequent frames using the wrong reference data, leading to persistent visual artifacts and temporal error propagation that lasts until the next independent INTRA frame is received.

The underlying idea behind ’764 is to decouple the tracking of reference frames from the standard temporal display order by implementing an independent numbering scheme. While traditional temporal references track the timing of every frame (including those not used for prediction), this invention introduces a dedicated sequence indicator that only increments for frames that serve as anchors for future prediction. By focusing exclusively on the sequence of reference pictures, the system provides a robust mechanism for a decoder to verify the integrity of its prediction chain regardless of how many non-reference frames were skipped or lost.

The claims of ’764 focus on a method and apparatus for assigning consecutive sequence indicator values to reference pictures in their encoding order. This numbering is strictly independent of the number of non-reference pictures (like B-frames) or non-coded pictures situated between them. The independent claims specifically cover the logic where a decoder identifies a gap in these values—detecting if the difference exceeds a predetermined amount—to immediately flag the loss or corruption of a critical reference frame, rather than relying on external transport-layer sequence numbers.

In practice, the encoder assigns a Reference Picture Order Number (RPON) to each I-frame or P-frame and embeds this value within the video bitstream, often utilizing Supplemental Enhancement Information (SEI) or picture headers. When the decoder receives a new reference frame, it compares the current RPON to the previously stored value. If the increment is greater than the expected step (typically one), the decoder instantly recognizes that a reference dependency has been broken. This allows the decoder to take proactive measures, such as freezing the display or signaling the encoder to transmit a new INTRA-coded recovery point.

This approach differentiates itself from prior art by solving the ambiguity inherent in standard temporal references and transport-layer sequence numbers. Traditional temporal references change based on the display timing, making it difficult to distinguish between a skipped non-reference frame and a lost reference frame. By using a dedicated reference counter, ’764 ensures that the decoder can maintain synchronization of the reference buffer even when variable encapsulation strategies or multi-layer scalability are employed, significantly improving visual stability in error-prone streaming and conferencing applications.