Intensity guided cost metric for fast stereo matching under radiometric variations.

Area of Science:

• Computer Vision
• Image Processing
• Machine Learning

Background:

• Stereo matching is crucial for 3D reconstruction but challenged by radiometric variations.
• Existing cost metrics often lack robustness or are computationally intensive.
• Radiometric changes between stereo images hinder accurate correspondence.

Purpose of the Study:

• To propose a novel similarity metric, the Intensity Guided Cost Metric (IGCM), for stereo matching.
• To enhance depth accuracy and reduce errors caused by radiometric variations.
• To develop a computationally efficient and robust stereo matching solution.

Main Methods:

• Introduced the Intensity Guided Cost Metric (IGCM) as a new similarity metric.
• Integrated IGCM with a color formation model to address radiometric variations.
• Utilized integral images for computational speedup.

Main Results:

• IGCM significantly improves depth accuracy by rejecting outliers and minimizing edge-fattening.
• Demonstrated substantial average error rate reductions compared to ANCC, DASC, AD, FCVF, and IGF methods on the Middlebury dataset.
• Achieved significant speedups (up to 45x) using integral images with IGCM.

Conclusions:

• The proposed IGCM is a robust and efficient solution for stereo matching under radiometric variations.
• IGCM offers superior performance in terms of accuracy and speed compared to existing methods.
• This metric advances the field of stereo vision, enabling more reliable 3D reconstruction.