Complex-Valued Disparity: Unified Depth Model of Depth from Stereo, Depth from Focus, and Depth from Defocus Based on...

Area of Science:

• Computer Vision
• Optics
• Image Processing

Background:

• Depth estimation from light fields is crucial for 3D scene understanding.
• Existing methods often struggle with non-Lambertian surfaces and require complex processing.
• A unified approach is needed to integrate various depth cues effectively.

Purpose of the Study:

• To propose a unified depth model for light fields using complex-valued disparity.
• To represent disparity in both Cartesian and polar coordinates for comprehensive analysis.
• To validate the model's efficacy in local disparity estimation.

Main Methods:

• Developed a unified depth model based on light field gradients.
• Represented estimated disparity as complex numbers.
• Analyzed disparity using real, imaginary, magnitude, and angle responses in 3D volumes.

Main Results:

• The complex-valued disparity model effectively captures depth information.
• The real part of disparity relates to in-focus plane estimation.
• The imaginary part quantifies non-Lambertian surface properties.
• Magnitude and real responses demonstrated validity for local disparity estimation.

Conclusions:

• The proposed complex-valued disparity model offers a unified framework for light field depth estimation.
• This approach enhances understanding of depth cues including stereo, focus, and defocus.
• The model shows promise for accurate local disparity estimation in challenging scenarios.