DCT-Former: Efficient Self-Attention with Discrete Cosine Transform.
Carmelo ScribanoGiorgia FranchiniMarco PratoMarko BertognaPublished in: CoRR (2022)
Keyphrases
- discrete cosine transform
- image compression
- dct coefficients
- transform domain
- dct domain
- filter bank
- coded images
- blocking artifacts
- lapped transforms
- compressed images
- image and video coding
- image blocks
- computationally efficient
- dc coefficients
- image processing
- discrete fourier transform
- spatial domain
- image coding
- multiresolution
- feature vectors
- face recognition