Digital Video Signal Modification TRIZ Cases

Specialized 8-bit and 10-bit LGT and DST cores extend AV1 transform options to improve video coding efficiency and accuracy.

Context-adaptive split flag decoding enables multi-tree picture partitioning, improving compression efficiency for non-square blocks.

Flag-based mixed NAL signaling lets subpictures change resolution independently, improving VVC decoding efficiency without excessive signaling overhead.

Spatial and temporal merge candidate rules cut memory access in inter prediction while improving cache hits and decoding speed.

Adaptive horizontal and vertical transform selection improves high-resolution video compression while limiting data volume.

Jointly deriving illumination compensation parameters from motion-compensated reference blocks improves bi-prediction accuracy and coding quality.

Ratio-based size information in an NNPFC SEI message lets decoders derive output picture dimensions with lower bitstream overhead and preserved image quality.

Preset bitstream identifiers let encoder and decoder switch or update AI models together, improving coding evolution with lower bitrate overhead.

Boundary-based sign prediction narrows transform coefficient sign candidates, cutting calculation load and residual derivation cycles in video decoding.

Resampling the neural-network post-filter input by picture size improves image coding efficiency while limiting high-resolution bitstream load.

Explicit NNPFC SEI signaling clarifies output picture generation, improving coding efficiency, decoder reliability, and image quality.

By matching sub-images across frames and reusing prior encoding data, this case cuts redundant desktop video encoding and bandwidth use.

Multiple linear parameter sets improve intra block copy illumination compensation, raising video prediction accuracy and coding efficiency.

Multiple template groups and a block vector candidate list improve Intra TMP prediction accuracy without exhaustive encoding search.

Selective optical flow hyperprior coding cuts video bitrate overhead by improving probability estimation for inter prediction and residual encoding.

Embedded suitability identifiers let decoders keep task-relevant images while reducing processing load and preserving execution accuracy.

Restricting picture types from NNPF SEI messages prevents encoder-decoder mismatches while improving bitstream transmission and storage efficiency.

Layered HDR encoding pairs base and enhancement data with scheme metadata to avoid transcoding exceptions and improve final synthesis.

Confidence values guide frame reconstruction when motion prediction is unreliable, reducing coding errors and improving image compression.

Neighbor-based intra prediction candidates improve image coding efficiency by lowering bit rates while preserving high image quality.

Conditional MPM list generation adds adjacent angular modes from neighboring blocks to improve multi-line intra prediction accuracy and coding efficiency.

Conditional syntax signaling of face video IDs and counts enables selective region extraction without unnecessary bitstream overhead.

Decoding a split flag and partition data lets pictures be divided into sub-units for parallel video processing with lower hardware load.

Recursive block and sub-block splitting handles image-edge blocks more efficiently, improving compression and neighboring block use.

Reduced suffix lengths for selected BVD codeword prefixes cut signaling overhead and improve video compression efficiency.

Metadata guides in-loop and post-filtering to a better spatial resolution, preserving edge detail and film grain after up-sampling.

A fixed maximum candidate count decouples index decoding from list construction, improving video decoder throughput and robustness.

Template matching selects MHP weights from block similarity, cutting video coding signaling overhead and bandwidth use.

Reusing indexed patch textures cuts residual signaling in screen content video, improving compression efficiency with lower memory and decoder load.

Conditional ACT signaling lets codecs switch color spaces within a sequence, improving coding efficiency for mixed screen and natural video.

Deblocking strength is adjusted by adjacent prediction modes so weighted intra-inter blocks suppress boundary distortion without added complexity.

Invertible geometric transforms realign distorted pixels and non-rectilinear blocks to improve intra-prediction and video compression efficiency.

Neighboring block modes guide weighted intra prediction, cutting transmitted mode data while preserving decoding accuracy and manageable complexity.

Correcting block motion vectors from integrated neighboring references improves prediction accuracy, cuts residuals, and reduces video bandwidth.

Neighbor and temporal block scanning builds padded boundary blocks for motion compensation, improving video compression at picture edges.

A classifier switches among specialized ML weights by video content and compression level to improve artifact removal and resolution upscaling.

Semantic information guides different filtering strengths across video frame regions to improve neural network accuracy and viewing quality.

Bit-depth-based clipping keeps ACT residual samples within range, improving video compression efficiency while preserving visual quality.

Stored inter prediction data removes intra prediction wait states, speeding video decoding across mixed coding unit sizes.

Gradient-based pixel-group motion compensation improves bi-directional video prediction accuracy while controlling decoding complexity.

Flag-based range-extension signaling lets image coders use bit-depth tools without breaking compatibility with decoders limited to standard ranges.

Curved intra prediction improves block-based image encoding by handling high-resolution content more efficiently and lowering transmission and storage costs.

Bilateral matching refines forward and backward motion vector differences to improve video compression in non-linear motion scenes.

Warped reference pictures improve inter prediction when global motion reduces frame similarity, preserving compression efficiency and image quality.

Pruning redundant HMVP motion candidates keeps merge and AMVP lists compact while preserving prediction accuracy and improving video compression.

Curved prediction modes and filtered reference samples improve intra block decoding efficiency while limiting added prediction complexity.

Multiple motion hypotheses and attention-weighted warped features improve temporal alignment and cut bit rate in learned video compression.

Block-level motion vector precision selection improves affine video prediction accuracy while limiting signaling overhead and bitrate.

Targeted in-loop filtering compensates mapping-induced signal changes between video blocks, improving coding efficiency and image quality.

A learned post-filter uses codec auxiliary information to improve decoded data for machine vision without embedding neural networks into the codec.