13.1: Invited Paper: Multigrid Backprojection Super‐Resolution and Deep Filter Visualization

  title={13.1: Invited Paper: Multigrid Backprojection Super‐Resolution and Deep Filter Visualization},
  author={Pablo Navarrete Michelini and Hanwen Liu and Dan Zhu},
  journal={SID Symposium Digest of Technical Papers},
We introduce a novel deep‐learning architecture for image upscaling by large factors (e.g. 4x, 8x) based on examples of pristine high‐resolution images. Our target is to reconstruct high‐resolution images from their downscale versions. The proposed system performs a multi‐level progressive upscaling, starting from small factors (2x) and updating for higher factors (4x and 8x). The system is recursive as it repeats the same procedure at each level. It is also residual since we use the network to… 

Multi–Grid Back–Projection Networks

A perceptual quality target aims to create more realistic outputs by introducing artificial changes that can be different from a high resolution original content as long as they are consistent with the low resolution input.

MGBPv2: Scaling Up Multi-Grid Back-Projection Networks

The second generation of MultiGrid BackProjection networks (MGBPv2) is introduced whose major modifications make the system scalable and more general than its predecessor, and which can balance between high quality and high performance.

Back–Projection Pipeline

This work proposes a novel solution to make back-projections run in multiple resolutions by using a data pipeline workflow and is represented by a system of ODEs, as opposed to a single ODE in the case of ResNets.

D2C-SR: A Divergence to Convergence Approach for Real-World Image Super-Resolution

D2C-SR is presented, a novel framework for the task of real-world image super-resolution with better accuracy and visual improvements against state-of-the-art methods, with a significantly less parameters number and the D2C structure can be applied as a generalized structure to some other methods to obtain improvement.

DIV8K: DIVerse 8K Resolution Image Dataset

The DIVerse 8K resolution image dataset (DIV8K) is introduced, which contains a over 1500 images with a resolution up to 8K, and is therefore the ideal dataset for training and benchmarking super-resolution approaches, applicable to extreme upscaling factors of 32x and beyond.

Multi-scale Recursive and Perception-Distortion Controllable Image Super-Resolution

A discriminator for adversarial training with the following novel properties is proposed: it is multi–scale that resembles a progressive–GAN; it is recursive that balances the architecture of the generator; and it includes a new layer to capture significant statistics of natural images.

Perception-oriented Single Image Super-Resolution via Dual Relativistic Average Generative Adversarial Networks.

Experimental results and ablation studies show that the proposed algorithm can rival state-of-the-art SR algorithms, both perceptually (PI-minimization) and objectively (PSNR-maximizing) with fewer parameters.

Inverse-Based Approach to Explaining and Visualizing Convolutional Neural Networks

A new inverse-based approach that computes the inverse of a feedforward pass to identify activations of interest in lower layers and develops a novel plot that shows the tradeoff between the amount of activations and the rate of class reidentification.

AIM 2019 Challenge on Image Extreme Super-Resolution: Methods and Results

This paper reviews the AIM 2019 challenge on extreme image super-resolution, the problem of restoring of rich details in a low resolution image, and gauges the experimental protocol and baselines for the extreme image Superresolution task.

Image super-resolution using progressive residual multi-dilated aggregation network

A progressive residual multi-dilated aggregation network (PRMAN), which performs multi-level upsampling to reconstruct images with large-scale factors and exceeds the state-of-the-art methods in most cases.



Deep Back-Projection Networks for Super-Resolution

It is shown that extending the idea to allow concatenation of features across up- and downsampling stages (Dense DBPN) allows us to reconstruct further improve super-resolution, yielding superior results and in particular establishing new state of the art results for large scaling factors such as 8× across multiple data sets.

Deep Laplacian Pyramid Networks for Fast and Accurate Super-Resolution

This paper proposes the Laplacian Pyramid Super-Resolution Network (LapSRN) to progressively reconstruct the sub-band residuals of high-resolution images and generates multi-scale predictions in one feed-forward pass through the progressive reconstruction, thereby facilitates resource-aware applications.

Enhanced Deep Residual Networks for Single Image Super-Resolution

This paper develops an enhanced deep super-resolution network (EDSR) with performance exceeding those of current state-of-the-art SR methods, and proposes a new multi-scale deepsuper-resolution system (MDSR) and training method, which can reconstruct high-resolution images of different upscaling factors in a single model.

Accelerating the Super-Resolution Convolutional Neural Network

This paper aims at accelerating the current SRCNN, and proposes a compact hourglass-shape CNN structure for faster and better SR, and presents the parameter settings that can achieve real-time performance on a generic CPU while still maintaining good performance.

Deep Image Demosaicking using a Cascade of Convolutional Residual Denoising Networks

This work proposes a novel deep learning architecture which is inspired by powerful classical image regularization methods and large-scale convex optimization techniques, and shows that the network has the ability to generalize well even when it is trained on small datasets, while keeping the overall number of trainable parameters low.

Image Super-Resolution Using Deep Convolutional Networks

We propose a deep learning method for single image super-resolution (SR). Our method directly learns an end-to-end mapping between the low/high-resolution images. The mapping is represented as a deep

A+: Adjusted Anchored Neighborhood Regression for Fast Super-Resolution

This work proposes A+, an improved variant of Anchored Neighborhood Regression, which combines the best qualities of ANR and SF and builds on the features and anchored regressors from ANR but instead of learning the regressors on the dictionary it uses the full training material, similar to SF.

Learning a Deep Convolutional Network for Image Super-Resolution

This work proposes a deep learning method for single image super-resolution (SR) that directly learns an end-to-end mapping between the low/high-resolution images and shows that traditional sparse-coding-based SR methods can also be viewed as a deep convolutional network.

Accurate Image Super-Resolution Using Very Deep Convolutional Networks

This work presents a highly accurate single-image superresolution (SR) method using a very deep convolutional network inspired by VGG-net used for ImageNet classification and uses extremely high learning rates enabled by adjustable gradient clipping.

Deeply-Recursive Convolutional Network for Image Super-Resolution

This work proposes an image super-resolution method (SR) using a deeply-recursive convolutional network (DRCN) with two extensions: recursive-supervision and skip-connection, which outperforms previous methods by a large margin.