ReNO: Enhancing One-step Text-to-Image Models through Reward-based Noise Optimization.
Luca EyringShyamgopal KarthikKarsten RothAlexey DosovitskiyZeynep AkataPublished in: CoRR (2024)
Keyphrases
- image noise
- input image
- background noise
- random noise
- multiscale
- image data
- image features
- image structure
- image statistics
- test images
- image content
- post processing
- image classification
- bayesian framework
- gaussian noise
- feature points
- single image
- geometric distortions
- degraded images
- web images
- image edges
- additive noise
- image segmentation
- denoising algorithm
- probabilistic model
- low frequency
- high contrast
- image retrieval
- image analysis
- high resolution
- image representation
- segmentation method
- signal to noise ratio
- random fields
- text mining
- end to end
- noise level
- reinforcement learning
- image processing
- image collections
- imaging process
- printed text