A 3D Convolutional Neural Network for Light Field Depth Estimation

Depth estimation has always been a great challenge in the field of computer vision and machine learning. There is a rich literature focusing on depth estimation in stereo vision or in monocular imaging, while the domain of depth estimation in light field images is still in its infancy. The paper pro...

Full description

Saved in:
Bibliographic Details
Main Authors: Faluvegi, Agota, Bolsee, Quentin, Nedevschi, Sergiu, Dadarlat, Vasile-Teodor, Munteanu, Adrian
Format: Conference Proceeding
Language:English
Subjects:
3D convolution
CNN
Computer vision
Convolution
Deep learning
depth estimation
disparity map
EPI
Estimation
Focusing
Light field image
Neural networks
Three-dimensional displays
Citations: Items that cite this one
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Depth estimation has always been a great challenge in the field of computer vision and machine learning. There is a rich literature focusing on depth estimation in stereo vision or in monocular imaging, while the domain of depth estimation in light field images is still in its infancy. The paper proposes a fully convolutional 3D neural network that estimates the disparity in light field images. The proposed method is parametric as it is able to adapt to input images of arbitrary size and it is lightweight and less prone to overfitting thanks to its fully convolutional nature. The experiments reveal competitive results against the state of the art, demonstrating the potential offered by deep learning solutions for disparity estimation in light field images.
ISSN:2379-1780
DOI:10.1109/IC3D48390.2019.8975996