Image Coding TRIZ Cases

Predicting boundary UV coordinates from prior mesh data cuts residual bitrate while preserving dynamic mesh reconstruction for real-time AR and VR.

Sparsity-based mode switching screens isolated nodes in point clouds, improving compression efficiency and reducing bitstream size.

A compact atlas hash in SEI messages lets decoders compare decoded atlas data and detect bitstream integrity or decoding errors.

Variable and fixed coefficient networks refine intra prediction signals to reduce residual data and improve coding efficiency.

A neural network with convolution and channel-wise attention reconstructs compressed LIDAR range and intensity data while reducing artifacts.

Capturing laser elevation and azimuth angles revise point coordinates to cut geometry distortion and improve point cloud coding efficiency.

Tracks point cloud slice data by unit type and adds ordering metadata so parsers can rebuild compliant bit-streams for decoding and rendering.

Separate point cloud media streams by geometry and attributes so receivers prioritize needed data under changing network conditions.

Grouped convolution and decoding indication information cut neural image decoding complexity, improving efficiency on mobile devices.

Linear correction parameters compensate quantization and decoding errors in neural image coding, improving reconstruction quality and rate-distortion.

Multiple quantization parameters for geometry, luminance, and chrominance improve point cloud coding efficiency while preserving decoding accuracy.

Visible and thermal image encoders map scenes to location pairs, enabling accurate agent self-localization when GPS and two-way links are unavailable.

Selecting nearby-point prediction modes lowers attribute residuals, reducing 3D data volume while preserving complete point cloud information.

Adaptive point sampling preserves key geometry during point cloud video compression, reducing redundancy while improving reconstruction quality.

A two-layer point cloud coding scheme signals local reconstruction modes to cut bit-rate while preserving reconstruction quality.

Threshold-based prediction in an N-ary point cloud tree improves attribute coding efficiency while avoiding unnecessary processing on sparse nodes.

Joint coding puts texture and 4:2:0 displacement data in one bitstream to cut encoder complexity and improve dynamic mesh coding efficiency.

NNPFC SEI messages let decoders apply neural-network post-filters across video sequences with less signaling overhead and lower bandwidth use.

To address inefficient signaling, NNPFC SEI messages specify neural-network post-filter purposes for picture rate upsampling and selective output.

Encoding cost and error metrics jointly prioritize edge collapses, improving progressive mesh rate-distortion performance with fewer bits at a given quality.

Intersection-aware vertex selection preserves mesh interactions during decimation, reducing visual errors in real-time 3D graphics rendering.

A neural network encodes 3D point-cloud super-points into latent vectors for robust object detection without labeled training data.

Hierarchical parent- and child-level encoding reduces shader executions during intersection testing while preserving sub-primitive presence data for ray tracing.

Separate ALF, scaling-list, and LMCS APS NAL units limit redundant signaling across pictures while preserving accurate slice decoding.

Parent-aware table selection separates n-ary tree nodes to reduce encoding and decoding workload while maintaining coding efficiency.

G-PCC and V-PCC workflows encode geometry and attributes for efficient transmission and decoding of detailed point cloud services.