Image super-resolution based on deep neural network of multiple attention mechanism

2021 ◽  
Vol 75 ◽  
pp. 103019
Author(s):  
Xin Yang ◽  
Xiaochuan Li ◽  
Zhiqiang Li ◽  
Dake Zhou
Keyword(s):  
Neural Network ◽  
Deep Neural Network ◽  
Super Resolution ◽  
Attention Mechanism ◽  
Image Super Resolution

RSAMSR: A deep neural network based on residual self-encoding and attention mechanism for image super-resolution

Optik ◽  
2021 ◽  
pp. 167736
Author(s):  
Xin Yang ◽  
Shiyu Wang ◽  
Jiali Han ◽  
Yingqing Guo ◽  
Tao Li
Keyword(s):  
Neural Network ◽  
Deep Neural Network ◽  
Super Resolution ◽  
Attention Mechanism ◽  
Image Super Resolution

Satellite image super-resolution based on progressive residual deep neural network

2020 ◽  
Vol 14 (03) ◽  
pp. 1
Author(s):  
Junwei Zhang ◽  
Shigang Liu ◽  
Yali Peng ◽  
Jun Li
Keyword(s):  
Neural Network ◽  
Deep Neural Network ◽  
Satellite Image ◽  
Super Resolution ◽  
Image Super Resolution

scholarly journals Image Formation Model Guided Deep Image Super-Resolution

2020 ◽  
Vol 34 (07) ◽  
pp. 11807-11814
Author(s):  
Jinshan Pan ◽  
Yang Liu ◽  
Deqing Sun ◽  
Jimmy Ren ◽  
Ming-Ming Cheng ◽  
...  
Keyword(s):  
Neural Network ◽  
Neural Networks ◽  
Deep Neural Network ◽  
Deep Neural Networks ◽  
Super Resolution ◽  
Image Formation ◽  
Formation Model ◽  
High Resolution Images ◽  
Image Super Resolution ◽  
Pixel Substitution

We present a simple and effective image super-resolution algorithm that imposes an image formation constraint on the deep neural networks via pixel substitution. The proposed algorithm first uses a deep neural network to estimate intermediate high-resolution images, blurs the intermediate images using known blur kernels, and then substitutes values of the pixels at the un-decimated positions with those of the corresponding pixels from the low-resolution images. The output of the pixel substitution process strictly satisfies the image formation model and is further refined by the same deep neural network in a cascaded manner. The proposed framework is trained in an end-to-end fashion and can work with existing feed-forward deep neural networks for super-resolution and converges fast in practice. Extensive experimental results show that the proposed algorithm performs favorably against state-of-the-art methods.

Sign in / Sign up

Export Citation Format

Share Document