Blurry-Edges: Photon-Limited Depth Estimation from Defocused Boundaries

Researchers at Purdue University have developed a novel approach to measure object depths from photon-limited images along defocused object boundaries. Extracting depth information from photon-limited, defocused images is challenging because depth from defocus (DfD) relies on accurate estimation of defocus blur, which is fundamentally sensitive to image noise. This innovative solution developed at Purdue overcomes traditional limitations to DfD by utilizing deep learning and a novel DfD relation. Their method is based on a new image patch representation, called Blurry-Edges, that stores and visualizes a rich set of low-level patch information, including boundaries, color, and blurriness. Using deep neural network architecture that predicts the Blurry-Edges representation, this method is capable of analytically calculating depth from low-quality, noisy images.

Technology Validation:

The researchers validated that the Blurry-Edges approach achieved the highest depth estimation accuracy on photon-limited images compared to other state-of-the-art DfD methods.

Advantages:

-Higher depth estimation accuracy

-Handles photon-limited or noisy images

-Leverages deep learning

Applications:

-Helps camera companies who want to robustly measure object depths from photon-limited images better than current state-of-the-art depth-from-defocus methods, which can't deal with high noise level well.

Publications:

A paper regarding this technology has been accepted to the IEEE/CVF Conference on Computer Vision and Pattern Recognition 2025.

List of accepted papers as of March 14th, 2025: https://cvpr.thecvf.com/Conferences/2025/AcceptedPapers

TRL: 4

Intellectual Property:

Keywords: Deep learning depth estimation,Low-light imaging solutions,Photon-limited image processing,Advanced boundary detection,Computational imaging algorithms,Noise-robust depth sensing,Depth-from-defocus enhancement,AI-powered imaging,High-accuracy 3D sensing,Camera depth measurement technology,Image noise reduction,Enhanced imaging for cameras,Depth mapping in challenging lighting,Neural network imaging models