Abstracts due 22 July
Browse the Call for Papers
>
30 January - 4 February 2027
San Francisco, California, US
Industry Event
Vision Tech 3D Session: Improving Stereo Camera Accuracy Through Enhanced Calibration and Depth Map Quality Metrics
21 January 2026 • 1:00 PM - 1:30 PM PST | West Expo Stage 2 (Moscone West, Exhibit Level) 
Stereovision systems are increasingly utilized in embedded applications, such as Automotive Driver Assistance Systems (ADAS), Agricultural Automation, or Industrial Robotics, where reliable 3D depth perception is crucial for navigation, object detection, and environmental understanding. However, the performance of these systems heavily depends on both the quality of their calibration and the accuracy of the depth maps they generate. Existing calibration approaches, such as those provided by OpenCV, often lack robustness, precision, and fail to achieve optimal performance due to embedded hardware and optics variability.

In this work, we present an enhanced laboratory calibration methodology specifically designed to improve the depth accuracy of stereovision systems. Our approach addresses key limitations of traditional methods by introducing an optimized testing setup and methodology, resulting in significantly improved distortion curves, Brown-Conrady model (OpenCV3, OpenCV5, and OpenCV12).

Beyond calibration, we introduce a dedicated benchmarking framework for assessing depth map quality in 3D vision systems. This framework includes metrics for evaluating both depth accuracy and hole artifact characteristics, two critical factors affecting the usability of 3D data in embedded scenarios. The depth accuracy metric quantifies the deviation of reconstructed depth from ground-truth references, while the hole analysis metric characterizes the spatial distribution and size of missing data regions.

We validate our approach on multiple stereo camera systems targeting automotive and agricultural robotics use cases, demonstrating measurable improvements in depth precision and robustness across various distances and lighting conditions. These results highlight the importance of precise calibration and holistic benchmarking for advancing the reliability of embedded 3D sensing systems.


Speaker

Laurent Chanas
 
 
Laurent Chanas
Image Science Director
DXOMARK IMAGE LABS (France)


Dr Laurent Chanas has been working in image quality assessment and improvement for more the 20 years. He is now head of R&D team of the image quality assessment solution, named Camera Analyzer, at DXOMARK IMAGE LABS. Previously Laurent worked at DXO Labs, developing image processing algorithms (several international patents) and managed the test and measurement laboratory. Laurent holds a master’s degree from “Ecole Nationale de l’Electronique et ses Applications” as well as a PhD thesis in Image Processing from Cergy-Paris University.

 


Event Details

FORMAT: Oral presentation followed by audience Q&A.
MENU: Coffee, decaf, tea and water will be available nearby.
SETUP: Theater seating.