EPINET: A Fully-Convolutional Neural Network Using Epipolar Geometry for Depth from Light Field Images.
Changha ShinHae-Gon JeonYoungjin YoonIn So KweonSeon Joo KimPublished in: CoRR (2018)
Keyphrases
- epipolar geometry
- light field
- feature points
- geometric constraints
- scene geometry
- scene depth
- convolutional neural network
- image pairs
- image correspondences
- camera motion
- input image
- epipolar lines
- epipolar constraint
- light field rendering
- fundamental matrix
- point correspondences
- image registration
- three dimensional
- object recognition
- affine epipolar geometry
- image based rendering
- multiple images
- image features
- depth images
- camera positions
- depth information
- stereo images
- depth map
- face detection
- single image
- defocus blur
- image formation
- surface normals
- camera pose
- structure from motion
- motion estimation
- image retrieval
- image sequences