Variable-length coding of dictionary-based literals and indices improves repeated-sequence compression while supporting real-time processing.
Sub-block division and map-based bitstream modes improve RGB image compression efficiency while preserving color image quality.
Dual frame memories split search areas into clusters to cut DRAM access latency and speed motion estimation in video encoding.
Biased scan orders for 8×4 and 4×8 transform blocks group non-zero coefficients better, improving entropy coding and lowering bit rate.
Vertical-biased scan patterns reorder 4×4, 8×4, and 4×8 transform coefficients to improve entropy coding in interlaced and progressive video.
Reordering data blocks with an affinity array shortens LZ offset encoding, improving compression for storage and transmission.
A programmable GPU decoding core combines shaders with VLD hardware to handle CABAC, CAVLC, and other video standards with lower latency.
Tree-structured finite-state debinarization cuts CABAC decoding load and storage needs while supporting real-time video throughput.
Dual PPS and CABAC monitoring let H.264 encoders switch to CAVLC when timing is missed, preserving real-time bitstream output.
A level/run-before encoder separates non-zero coefficients and zero runs to cut CAVLC pipeline breaks and improve parallel processing.
Joint sparsity and common-component measurements reconstruct correlated signals with fewer samples and lower sensor communication costs.
Separating MSB and LSB data in reference frames cuts storage and bandwidth while preserving image quality through adaptive coding.