Image Enhancement TRIZ Cases

A 3D point cloud modeling method fuses feature-based pose data with direct-method depth estimates to generate dense spatial maps.

An inspection support device uses trained models to analyze generator images captured by robots.

Regional image segmentation and key point alignment reduce data processing time while maintaining high recognition accuracy.

A detection device calculates reliability information to generate precise three-dimensional shape models from single viewpoint data.

Separates hue and brightness via color space conversion to reduce pattern interference during capsule defect detection.

Image processing apparatus classifies intraluminal image areas to calculate accurate three-dimensional pixel values.

A stereo camera pair captures images from opposite ends of an airframe member to quantify structural deformation and strain through image analysis.

A method divides mask images into regular geometrical patterns to define landmarks for precise image registration.

Oscillator array transmits ultrasonic waves through a fixing tool that retains the specimen via a drive mechanism, resolving interference from clip holders.

A medical information processing device generates volume rendering images from X-ray CT data and superimposes color surface images onto them.

Encoding binary data into screen images allows terminals to exchange files without radio modules, eliminating electromagnetic interference.

Synthesizing a semi-transparent blue QR code resolves the conflict between large marker visibility and background occlusion in augmented reality applications.

A digital projection system guides lumber truss assembly using calibrated light and dynamic image adjustment.

Processed point cloud data simulates environmental changes for autonomous vehicle positioning system testing.

Merging depth and image modalities resolves detection accuracy versus system complexity trade-offs in retail analytics.

Generating pseudo-labeled data from weakly-labeled inputs expands training sets, improving prediction accuracy and generalization.

A self-supervised depth estimation method predicts depth maps from vehicle-mounted camera images using iterative warping and projection.

Segmenting point clouds into sectors reduces processing time while maintaining feature matching accuracy for autonomous vehicle navigation.

A behavior recognition system segments multiple workers using time series skeleton data and position coordinates to isolate the intended subject.

Extends color channels via boundary detection to prevent image distortion from mis-registration in printers.

Sequential matrix camera scans with pulsed illumination resolve reflective interference, lowering power consumption while maintaining inspection reliability.

A preoperative planning system generates bone graft templates using image processing to guide surgical extraction.

Applying gamma correction to correct and output images during neural network learning prevents gradation collapse in high luminance parts.

Combines magnetic resonance spatial data with microwave dielectric measurements to identify tissue boundaries.

A biological substance quantification method corrects fluorescent dye particle counts using reference luminance values from standard samples.

Synchronized multi-camera imagery feeds neural classifiers to detect body parts and construct probabilistic graphical models for pose recognition.

An adaptive subsampled look-up table reduces memory bandwidth usage in camera systems by storing varying levels of correction data across image patches.

Front and back cameras capture out-of-viewing-angle information to resolve the contradiction between simple operation and investigation accuracy.

Fusing laser radar and sonar data with inertial navigation for precise unmanned surface vehicle positioning.

Fine-tuned face identification model generates embeddings to cluster tracks without manual labeling.

Laser-induced electron tunneling enables voltage contrast signals through sacrificial oxide layers in deep features.

A system generates personalized 3D body models from standard 2D images captured by portable devices.

A calculation unit determines a target resolution value to generate recovery filters for consistent image quality.

An image processing system normalizes and co-registers ocular fundus images to synthesize temporal differences.

Low-coherence quantitative phase microscopy maps depth-resolved optical path-length differences to quantify nanoscale nuclear architecture in unstained tissue.

A segmentation network uses user corrections to overtrain on specific frames for accurate scene processing.

Corrects warped edges in depth maps using local error metrics to resolve edge accuracy deterioration caused by DIBR warping.

Integrating a time of flight system with an image sensor eliminates multi-camera complexity and post-processing artifacts during depth of field adjustments.

A processing device calculates three-dimensional displacements from time-series images to estimate supporting member motion.

Grid code mapping resolves monitoring complexity by projecting multi-camera data onto a single reference plane for unified object tracking.

A gaze locking detection system generates feature vectors from eye region pixel values to identify direct eye contact.

Image processing identifies raised hands for accurate survey counts, replacing manual counting that causes time loss and errors.

Region-specific correction coefficients adjust object size values for lens distortion, reducing computational complexity while maintaining detection accuracy.

Computational deconvolution replaces polarizing filters to resolve measurement accuracy losses in low reflectivity regions caused by high reflectivity glare.

A trained model segments input images to create determination labels for automatic training label correction.

Mounting reflectors on a moving vehicle enables automated camera calibration without road closures, resolving traffic disruption and motorist distraction.

Convolutional neural networks generate face feature vectors for subject identification in digital images.

A textured digital elevation model generation method projects lidar points into texel swaths using normalized coordinates and enhanced output pixel mapping.

A detail grade map controls adaptive spatial denoising power across image frames.