State of the Art

Image Denoising : In [10], Cheng et al. proposed to use patch guided internal clustering algorithm for image denoising. It utilizes Gaussian mixture model learning to guide the clustering of noisy images followed by an approximation process to estimate the subspace for image recovery. Zhang et al. [11] introduced the concepts of short-term and restricted long-term memory by making use of skip connections to pass information through the layers of the network.

Single Image Super Resolution : Mao et al. [12] introduced symmetric skip connections into a 30-layer convolutional auto-encoder network for image denoising and single image super resolution.

JPEG Deblocking : In [13], Dong et al. introduced an extended convolution network called Artifacts Reduction Convolutional Neural Networks (ARCNN) for removing the JPEG compression artifacts effectively.