NLIP: Noise-robust Language-Image Pre-training.
Runhui HuangYanxin LongJianhua HanHang XuXiwen LiangChunjing XuXiaodan LiangPublished in: CoRR (2022)
Keyphrases
- image noise
- salt pepper
- geometric distortions
- image data
- image features
- image structure
- additive noise
- single image
- input image
- random noise
- image classification
- image edges
- image analysis
- image retrieval
- image transformations
- high contrast
- multiscale
- image content
- image regions
- high resolution
- background noise
- segmentation algorithm
- partial occlusion
- geometric attacks
- intensity gradient
- test images
- training and testing data
- robust statistical
- degraded images
- programming language
- denoising algorithm
- noisy environments
- low frequency
- pixel values
- high frequency
- image representation
- image watermarking
- segmentation method
- feature vectors
- noise level
- weak edges
- imaging artifacts
- image processing