Noise or Signal: The Role of Image Backgrounds in Object Recognition.
Kai Yuanqing XiaoLogan EngstromAndrew IlyasAleksander MadryPublished in: ICLR (2021)
Keyphrases
- random noise
- object recognition
- image features
- additive noise
- template matching
- image noise
- image representation
- image structure
- low frequency
- low signal to noise ratio
- transformed image
- input image
- background noise
- image data
- white noise
- noise variance
- power spectrum
- image segmentation
- image content
- computer vision
- low snr
- pixel intensities
- high frequency
- single image
- multiscale
- image retrieval
- image classification
- bandpass
- quantization noise
- additive gaussian noise
- wiener filter
- nonlinear filters
- natural images
- denoising methods
- removing noise
- visual object recognition
- noise reduction
- multiplicative noise
- image edges
- high resolution
- d objects
- edge detection
- frequency domain
- signal processing
- noisy images
- denoising algorithm
- power spectra
- signal noise ratio
- noise model
- noise removal
- image enhancement
- contrast enhancement
- noise level
- geometric distortions
- response function
- morphological filters