Zobrazeno 1 - 5
of 5
pro vyhledávání: '"Janjušević, Nikola"'
Nonlocal self-similarity within images has become an increasingly popular prior in deep-learning models. Despite their successful image restoration performance, such models remain largely uninterpretable due to their black-box construction. Our previ
Externí odkaz:
http://arxiv.org/abs/2407.18967
Nonlocal self-similarity within natural images has become an increasingly popular prior in deep-learning models. Despite their successful image restoration performance, such models remain largely uninterpretable due to their black-box construction. O
Externí odkaz:
http://arxiv.org/abs/2306.01950
Image processing neural networks, natural and artificial, have a long history with orientation-selectivity, often described mathematically as Gabor filters. Gabor-like filters have been observed in the early layers of CNN classifiers and even through
Externí odkaz:
http://arxiv.org/abs/2204.11146
CDLNet: Noise-Adaptive Convolutional Dictionary Learning Network for Blind Denoising and Demosaicing
Deep learning based methods hold state-of-the-art results in low-level image processing tasks, but remain difficult to interpret due to their black-box construction. Unrolled optimization networks present an interpretable alternative to constructing
Externí odkaz:
http://arxiv.org/abs/2112.00913
Deep learning based methods hold state-of-the-art results in image denoising, but remain difficult to interpret due to their construction from poorly understood building blocks such as batch-normalization, residual learning, and feature domain proces
Externí odkaz:
http://arxiv.org/abs/2103.04779