ResFPN: Residual Skip Connections in Multi-Resolution Feature Pyramid Networks for Accurate Dense Pixel Matching.
RishavRené SchusterRamy BattrawyOliver WasenmüllerDidier StrickerPublished in: CoRR (2020)
Keyphrases
- multiresolution
- dense disparity map
- coarse to fine
- matching cost
- wavelet transform
- stereo matching
- stereo correspondence
- wide baseline stereo
- feature correspondences
- image pyramids
- matching algorithm
- pattern matching
- point to point correspondences
- subband
- input image
- laplacian pyramid
- matching process
- image pixels
- spatial pyramid matching
- densely sampled
- stereo matching algorithm
- hierarchical representation
- disparity map
- image fusion
- graph matching
- network structure
- keypoints
- image features
- feature vectors
- multiscale
- social networks
- preprocessing stage
- pixel values
- dense correspondence
- image pyramid
- high quality