Improving Realistic Worst-Case Performance of NVCiM DNN Accelerators through Training with Right-Censored Gaussian Noise.
Zheyu YanYifan QinWujie WenXiaobo Sharon HuYiyu ShiPublished in: CoRR (2023)
Keyphrases
- gaussian noise
- worst case
- training process
- noisy images
- denoising
- image restoration
- noise level
- noise model
- signal to noise ratio
- white noise
- noise removal
- median filtering
- impulsive noise
- noise variance
- impulse noise
- jpeg compression
- noise suppression
- salt and pepper noise
- lower bound
- np hard
- multiresolution
- training set
- training data
- image denoising
- binary images
- ground truth
- computational complexity