Image Enhancement TRIZ Cases

This case encodes geometry and color data for efficient 3D transmission, then analyzes user input to adapt displayed content.

A camera and orientation data establish a shared reference frame, reducing mapping effort and marker dependence during device alignment.

Multiple images with varying illumination are compared by pixel intensity change to select the least motion-affected corrected image.

Depth-data tracking uses unscented Kalman filters and adaptive model likelihoods to estimate target position during motion and rest.

Camera images and an autoencoder identify filaments leaving the guide, improving yarn path abnormality detection.

Automated shelf imaging checks product placement and empty spaces for consistent compliance.

This case adapts animal shape analysis across imaging directions to estimate weight without requiring the animal to stand on a scale.

A crawler-based radiography system distinguishes pipe insulation from the wall to map corrosion and erosion across multiple locations.

This case uses gradient thresholds, Laplace and discrete cosine transforms to remove glass ghosting from a single image.

A high-pass comparison and frequency-domain loss help CNN denoising preserve fine detail while improving SNR and SSIM.

Coregistered base and informer images generate clearer labels for machine learning detection and classification of biological structures.

This case uses patient-table keypoint detection to register optical and medical imaging coordinates without large calibration phantoms.

Visual detection, database control, and separable tool units support flexible, real-time weed treatment across varied carriers.

This case maps camera and RADAR/LIDAR object trajectories into a shared latent space to resolve coordinate and timing mismatches.

Independent image features are pooled symmetrically to expand HDR range while avoiding motion-related frame alignment.

Multi-spectral, multi-exposure imaging and verification networks pre-screen serum or plasma integrity, reducing errors and specimen waste.

This image-processing approach expands inner cell regions to separate crowded cells and measure each cell in multiplex-stained specimens.

This case fuses image and non-image features through product correlations, supporting accurate prediction of future or past states.

Automatic clustering and intensity checks locate calibration standards in unstructured lidar clouds for field-ready sensor alignment.

This case adapts webcam vital sign parameters to cohort statistics, reducing noise and tracking individual physiological changes.

Uncertainty-guided sampling supports clear medical rendering while avoiding preprocessing latency.

X-ray Talbot images provide accurate fiber orientation data for structural analysis without destructive sampling of large composites.

This case combines total-station prism tracking with multi-view images to map heavy equipment work in 3D during operation.

Motion vectors and image blending improve vehicle camera image reproduction without synchronizing to LED light sources.

Pixel-based quality metrics discard unreliable images, improving surface condition estimates for real-time agricultural operations.

Object detection and segmentation blur non-work-related areas in a see-through AR display, preserving task focus.

Imaging-specific extraction coefficients speed pigment concentration estimates.

Detecting signal polarity enables fast WASAB1 fitting of B0 and B1, producing accurate maps for MRI field correction.

Machine learning detects bad telematics installations from fastener orientation.

A trained teacher network guides monocular depth learning, helping vehicles recover scene scale without LiDAR at deployment.

Detect deformation abnormalities during medical image registration for more accurate alignment.

Depth sensors and pose-aware alignment combine partial scans into a full-body 3D model with accurate body measurements.

This evaluation system weights false estimation types by impact for more relevant tracking algorithm performance assessment.

To address VR-to-AR incompatibility, computational modules extract depth information and reformat content into multifocal virtual images.

This case separates vessel segmentation from CNN classification to limit information loss in intracranial occlusion imaging.

Two-pass skin filtering preserves realistic facial texture during stretching and compression.

Multiple detection criteria and selective user correction use reference images to improve infrastructure defect inspection accuracy.

Orthogonal polarization reduces surface glare while infrared light improves depth visibility for non-invasive lymph node and vessel imaging.

Automatic thumbnails track standard anatomical planes and reduce scanning workload.

A conversion model turns pixel distances into physical track measurements, enabling alerts for rail gaps and surface defects.

This case trains an inpainting neural network to separate same-label object instances while reducing manual guidance.

Preview blur levels before capture for intuitive background shading.

Track gaming activity despite obstructions with calibrated multi-camera vision.

This case uses machine learning to segment clothing and apply facial-expression-driven AR effects without depth sensors.

Adjusting scan timing across frames and synthesizing distance images improves resolution and range while preserving S/N ratio.

Generate diverse annotated virtual humans to train robust keypoint models.

Correct wide-angle thermal distortion to locate people accurately indoors.

A second PET volume during radiotracer bolus transit bridges functional PET data and CT angiography for accurate registration.

This case uses adaptive persistence and infinite impulse response filtering to reduce flash artifacts in low-velocity power Doppler imaging.

A camera-guided light unit activates device groups for relevant image regions, improving depth extraction while limiting power use.