【Scope of Work】

・Objective Tuning

- Quantification of Image Quality and Target-Based Optimization
Quantified image quality using charts and measurement equipment in controlled and calibrated lab environments
Optimized parameters based on key metrics such as SNR, dynamic range, color reproduction (ΔC), and resolution (MTF50) to meet defined target values

- Baseline Establishment
Optimized fundamental parameters including AE (Auto Exposure), AWB (Auto White Balance), CCM (Color Correction Matrix), and gamma correction
Established a stable baseline prior to transitioning to tuning phases under driving conditions

・Subjective Tuning

- Scene-Based Image Optimization Under Driving Conditions
Established a feedback loop for scene-based image evaluation and tuning under diverse driving scenarios, such as dusk, forest roads, and underground parking facilities
- Dynamic Range Optimization
Suppressed highlight clipping and black crush in extreme contrast scenes, such as tunnel entrances and exits
Fine-tuned LTM parameters to reduce temporal luminance fluctuations (flicker), achieving stable tonal reproduction for skies and road surfaces
- AWB Optimization Under Mixed Lighting Conditions
Adjusted AWB algorithms to prevent overcompensation for specific light sources
Reduced color casts (yellow/blue) in mixed-light environments, such as sodium lamps inside tunnels combined with natural daylight, or LED streetlights mixed with storefront lighting at night
- LED Flicker and Artifact Mitigation
Implemented countermeasures for LED flicker in traffic signals and road signs
Reduced ghosting and flare artifacts to minimize false detections in recognition algorithms
- Edge Enhancement and Noise Reduction
Balanced ISP parameters to suppress noise while preserving high-frequency details such as text on traffic signs and vehicle edges
Improved recognition accuracy under nighttime and adverse weather conditions

Practical Example 1: Image tuning for edge enhancement and noise reduction to improve visibility of road signs and roadside vegetation

Practical Example 2: Scene-based image tuning to emphasize color differences, enabling AD sensing systems to more reliably distinguish brake lamps from tail lamps