SMUDLP: Self-Teaching Multi-Frame Unsupervised Endoscopic Depth Estimation with Learnable Patchmatch.
Shuwei ShaoZhongcai PeiWeihai ChenXingming WuZhong LiuZhengguo LiPublished in: CoRR (2022)
Keyphrases
- multi frame
- depth estimation
- super resolution
- depth map
- motion estimation
- stereo vision
- image sequences
- low resolution
- high resolution
- stereo matching
- depth information
- point correspondences
- scene understanding
- single image
- dynamic scenes
- image pairs
- feature matching
- motion analysis
- optical flow
- stereo pair
- high quality
- d scene
- computationally efficient
- filtering algorithm
- specular reflection
- disparity map
- semi supervised
- real scenes
- training data
- three dimensional
- optic flow
- motion segmentation
- computer vision
- motion vectors
- viewpoint