Image Analysis TRIZ Cases

Two-stream convolutional neural network learns foreground and background feature embeddings for digital image search.

Segmented sub-projection units achieve wide viewing angles without large reflectors, reducing device volume and manufacturing costs.

A calibration method adjusts flexible display alignment by activating reserved pixel rows to substitute for shifted regions.

Spectral ratio analysis distinguishes material edges from shadows, reducing false positives in varying lighting conditions.

A semiconductor apparatus detects backside Z height double derivative distribution to align wafers at the ideal positioning center during thin film processing.

An enclosed chamber with a pH-sensitive substrate enables continuous mold detection through real-time optical sensing.

A verification method calculates geometric coordinates relative to a character centroid for display validation.

A determination assistance unit processes images of a specimen aspirating unit to remove visual obstructions.

A comparison unit matches X-ray images against reference data to identify hazardous objects.

A thin plate spline transform estimates non-rigid object motion by minimizing error between a template and back-warped pixels.

A feature amount mixing unit projects image and text vectors onto a shared space to generate mixed features.

A rendering system samples color contribution values selectively based on neighbor data to reduce computational load.

A motion video camera system calculates object size relative to line distance for accurate crossing detection.

Transect plotting and averaging reduce noise while capturing structural anisotropy, resolving instability from directional sensitivity.

Shadow detection verifies boundary positions against vertical noise and brightness variations to improve measurement precision.

An image capturing system generates custom dictionary data for subject detection using neural network constraints.

Recurrent neural networks select video frames via reinforcement learning to reduce processing time while maintaining detection accuracy.

A computer system presents visual performance feedback for a data-analysis pipeline to enable interactive modification of machine-learning operations.

A game map boundary rendering system expands a fuzzy boundary inward along normal directions to form a grid structure for visual output.

A landmark database associates facial part landmarks with state information to generate characteristic face data for virtual humans.

A double-column deep convolutional neural network merges global and local image representations to classify visual features.

Motion-compensated Bayesian estimation merges spatial and temporal information to reduce blocking artifacts during video upsampling.

Interleaved imaging pulses guide pushing ultrasound waves to reposition stones, reducing tissue damage and recurrence rates.

Controllable lights cast shadows on turbine components to reveal surface defects through automated imaging analysis.

A commodity reading apparatus calculates feature value similarity before learning processing to prevent misclassification of new items.

Remote systems analyze video data to identify stationary and moving objects within a camera field of view, then select customized encoding parameters.

Projects 3D volume data onto 2D planes to resolve contradictions between segmentation speed and consistency while maintaining structural accuracy.

A mobile camera system localizes optical patterns to zoom in on regions of interest for faster decoding.

Standardized SDP attributes transmit camera calibration data to resolve stitching precision issues while reducing remote data transportation volume.

A template-based information processing device associates image regions with analysis processes to extract and classify target areas without continuous learning.

A Field Programmable Gate Array performs trilinear interpolation on mapped color coordinates to generate corrected pixel data.

A 3D area-scan camera captures images at known intervals to extract object data.

Optimizing sensor positions on the face reduces device complexity while maintaining recognition accuracy in noisy environments.

A virtual prop detection method displays perspective information for obstructed items within a visual angle range.

Elastic block-wise image comparison ignores resolvable post-production alterations to detect unresolvable content changes in advertisements.

A computer generates a grayscale image by weighting color components based on glasses characteristics to detect eyes.

A mobile device generates diminished reality virtual images of retail shelves to visualize product replacements and optimize planograms.

A face detecting unit guides background model creation using depth data to isolate human body regions from moving objects.

An information processing device stores object data with positional coordinates to enable display functionality.

Machine learning classifiers identify segmented patches in distorted structured light patterns, enabling accurate triangulation despite noise and imperfections.

System calculates appearance probability based on movement trajectory to predict the next camera view before a subject disappears from current frame.

A conic section calibration object projects images of all possible calibration points using curved surfaces to enable precise parameter deduction.

An alert system adjusts notification strength based on driver face orientation to improve hazard awareness.

A 3D structured light scanner computes surface point clouds to calculate feature volumes without physical contact.

An object recognition apparatus recalculates similarity degrees between newly captured images and stored reference data to resume processing automatically.

Encoding apparatus assigns depth fidelity parameters to multi-view frames, resolving rendering quality deterioration from inconsistent depth information.

Segmentation and parameter changes enable accurate tracking of COPD progression by comparing actual tissue against individual baselines.

Evaluates 48 octahedral group transformations on orbit-responsive representations to resolve alignment speed versus accuracy trade-offs.