Bframes

B-frames, short for bidirectionally predictive frames, are a type of inter-frame video frame used in various video compression standards such as MPEG-2, H.264/AVC, and HEVC. They achieve higher compression by using information from both prior and future frames to predict the current frame, rather than relying on past frames alone as P-frames do or no prediction as I-frames.

How they work: For each B-frame, the encoder draws predictions from two reference frames surrounding it in display order: a past reference frame and a future reference frame (both typically I- or P- frames). It performs motion estimation against both references, generates two block predictions, and combines them (bi-prediction). The residual between the actual block and the combined prediction is encoded. Because the future reference frame is required, B-frames introduce a display-to-encoding order gap: frames must be buffered and encoded after the next reference frame has been reconstructed, leading to reordering.

Impact and trade-offs: B-frames generally reduce bitrate and improve quality for a given objective by exploiting temporal redundancy more effectively than I- and P-frames alone. However, they increase encoding and decoding complexity, require more memory for buffering, and add latency, which can be an issue for real-time applications and streaming error resilience.

Usage: The number of consecutive B-frames between anchor frames varies by codec and profile; common configurations include one or more B-frames between I/P references. Their presence is particularly beneficial in high-motion or high-resolution content when bitrate is constrained.