M4Depth: A motion-based approach for monocular depth estimation on video sequences.
Michaël FonderDamien ErnstMarc Van DroogenbroeckPublished in: CoRR (2021)
Keyphrases
- depth estimation
- depth cues
- image sequences
- video sequences
- dynamic scenes
- depth map
- monocular images
- depth estimates
- depth perception
- space time
- stereo vision
- motion estimation
- depth information
- camera motion
- optical flow
- ego motion
- moving objects
- stereo matching
- moving camera
- motion segmentation
- super resolution
- real scenes
- human motion
- dense depth maps
- object tracking
- background subtraction
- motion field
- three dimensional
- motion model
- multi view
- stereo pair
- d scene
- spatio temporal
- frame rate
- computer vision
- depth from defocus
- video data
- motion vectors
- motion blur
- disparity map
- structure from motion
- depth data
- surface reconstruction
- object recognition
- motion parallax
- object boundaries
- video frames
- foreground objects
- multi camera
- depth images
- high quality
- stereo images