Image Data Processing Details TRIZ Cases
- Multi-View Rendering of Non-Diffuse Objects Under Bandwidth Limits
Selective transmission of non-diffuse pixel data preserves reflective surface rendering in multi-view video while limiting bandwidth and compute load.
- Image Conversion Model Training with Region-Focused Patch Loss
Region-focused patch sampling improves image conversion training by emphasizing specific objects, raising cross-environment accuracy for vehicles.
- Encrypted Screen Watermarking for Unauthorized Photo Traceability
A time-varying encrypted screen mark captures device ID and timestamp in photos, enabling traceability of unauthorized screen photography.
- Video Watermark Segmentation for Robust Extraction After Compression
Segmented video watermarking embeds multiple codes across frame sequences to preserve ownership detection after compression, clipping, and frame-rate changes.
- Video Object Attribute Watermarking for Robust Leak Tracing
User-specific changes to background object attributes make video watermarks harder to detect, more attack-resistant, and traceable.
- Watermarked Media Timestamp Sync for Delay-Accurate Client Actions
Embedded timestamps let media clients compute clock offset from received content, compensating for transmission delay in timed actions.
- Fingerprint Imaging Row Driver Circuit for Faster Authentication
Variable scan-signal rates in a fingerprint imaging row driver cut authentication time while preserving detection accuracy.
- Image Watermark Embedding Using Color-Channel Frequency Complexity
Frequency-based color-channel analysis varies watermark strength by image area to improve extraction on unseen images while preserving quality.
- 6DoF Video Playback Using Line-of-Sight View Stream Selection
Line-of-sight stream selection lets 6DoF video switch viewpoints in virtual 3D space without relying on one fixed viewing position.
- Dual Watermark Embedding for Model-Independent Source Tracing
Dual watermarks link generated content to both user and model sources without retraining, enabling faster extraction without model access.
- Blockchain Watermark Allocation for Decentralized Media Measurement
A private blockchain allocates and verifies watermark payloads to reduce human error, prevent duplication, and support near-real-time media measurement.
- Neural Visibility Modeling for Imperceptible Watermark Masking
A JND-trained neural visibility model predicts local masking thresholds to keep watermarking imperceptible while preserving robustness and capacity.
- Thin Client Graphics Rendering With AI-Based Real-Time Setting Prediction
AI predicts real-time graphics settings for thin clients to balance visual quality, CPU/GPU load, and network constraints.
- Multiple Digital Watermarking With Orthogonal Frequency Embedding
Orthogonal frequency-domain watermarks avoid mutual interference, enabling independent detection and stronger media traceability.
- Watermark Metadata Access for Secure Multimedia Database Population
Embedded watermarks preserve multimedia metadata through transcoding and transmission, enabling secure lookup and database population with less bandwidth.