On the Synergies Between Machine Learning and Binocular Stereo for Depth Estimation From Images: A Survey

Area of Science:

• Computer Vision
• Machine Learning
• Deep Learning

Background:

• Stereo matching is a foundational computer vision problem with a 40-year research history.
• The field has evolved from local pixel-level methods to optimization techniques and now data-driven approaches.
• Deep learning has recently revolutionized stereo matching, enabling novel applications.

Purpose of the Study:

• To review recent advancements in learning-based depth estimation using single and binocular images.
• To highlight the synergistic relationship between machine learning and stereo vision.
• To identify current successes and future challenges in the field.

Main Methods:

• Review of recent research in learning-based depth estimation.
• Analysis of deep learning methodologies applied to stereo matching.
• Examination of techniques for self-supervised monocular depth estimation.

Main Results:

• Deep learning has significantly advanced the state-of-the-art in stereo matching.
• Stereo vision techniques have enabled new deep learning-based depth estimation methods.
• Successful applications of learning-based depth estimation from single and binocular images have emerged.

Conclusions:

• The integration of deep learning has transformed stereo matching and depth estimation.
• Significant progress has been made, but open challenges remain in learning-based depth estimation.
• Future research will likely focus on further enhancing these synergistic approaches.