Super Resolution

Super Resolution

Super Resolution is the process of recovering a High Resolution (HR) image from a given Low Resolution (LR) image. It is an important class of image processing techniques in computer vision and image processing and enjoys a wide range of real-world applications, such as medical imaging, satellite imaging, surveillance and security, and astronomical imaging, amongst others.

With the advancement of deep learning techniques in recent years, deep learning-based SR models have been actively explored and often achieve state-of-the-art performance on various benchmarks. A variety of deep learning methods have been applied to solve SR tasks, ranging from the early Convolutional Neural Networks (CNN) to recent cascading residual architectures.

Our team is working on the optimization of a recent state-of-the-art model, the Densely residual Laplacian Network. We developed a simplified version of this model that reduces memory consumption and increases time efficiency without sacrificing the accuracy and providing comparable results. We have thus shown that the task of SR can be efficiently achieved with less complex deep learning architectures.


Related Papers

2023

Asfa Jamil and Alessandro Artusi

Computational efficient deep learning-based super resolution approach

Deep learning-based Single Image Super-Resolution (SISR) has recently provided great performances when compared with state-of-the-art approaches. However, these performances are usually at the cost of high computational complexity and memory management, even at the inference stage. In this paper, we aim to reduce the structural complexity of a state-of-the-art Deep Neural Networks (DNN)1 approach to propose a cost-effective solution to the problem of SISR. We have investigated how the different components of the model (baseline) may affect the overall complexity while minimizing the negative effect on its quality performances. This has provided a solution, which yields quality performances comparable to the baseline model, while approximately reducing more than one order of magnitude the number of parameters, the spatial complexity (GPU memory) up to 1/6 and inference time by 1/2.

Close

@inproceedings{10.1117/12.2665645,
author = {Asfa Jamil and Alessandro Artusi},
title = {{Computational efficient deep learning-based super resolution approach}},
volume = {12571},
booktitle = {Real-time Processing of Image, Depth and Video Information 2023},
editor = {Matthias F. Carlsohn},
organization = {International Society for Optics and Photonics},
publisher = {SPIE},
pages = {125710E},
keywords = {super resolution, deep learning, efficient computational performances, efficient memory management, real-time performances, real-time, image processing, image/video encoding},
year = {2023},
doi = {10.1117/12.2665645},
URL = {https://doi.org/10.1117/12.2665645}
}

Close

2022

A. Basso, P. Ribeca, M. Bosi, N. Pretto, G. Chollet, M. Guarise, M. Choi, L. Chiariglione, R. Iacoviello, F. Banterle, A. Artusi, F. Gissi, A. Fiandrotti, G. Ballocca, M. Mazzaglia and S. Moskowitz

AI-Based Media Coding Standards

Moving Picture, Audio and Data Coding by Artificial Intelligence (MPAI) is the first standards organization to develop data coding standards that have artificial intelligence (AI) as their core technology. MPAI believes that universally accessible standards for AI-based data coding can have the same positive effects on AI as standards had on digital media. Elementary components of MPAI standards–AI modules (AIMs)–expose standard interfaces for operation in a standard AI framework (AIF). As their performance may depend on the technologies used, MPAI expects that competing developers providing AIMs will promote horizontal markets of AI solutions that build on and further promote AI innovation. Finally, the MPAI framework licences (FWLs) provide guidelines to intellectual property right (IPR) holders facilitating the availability of compatible licenses to standard users.

Close

@article{basso2022ai,
title={AI-Based Media Coding Standards},
author={Basso, Andrea and Ribeca, Paolo and Bosi, Marina and Pretto, Niccol{\`o} and Chollet, G{\’e}rard and Guarise, Michelangelo and Choi, Miran and Chiariglione, Leonardo and Iacoviello, Roberto and Banterle, Franesco and others},
journal={SMPTE Motion Imaging Journal},
volume={131},
number={4},
pages={10–20},
year={2022},
publisher={SMPTE}
}

Close

2021

L. Chiariglione, et al. (A. Artusi)

AI-based Media Coding and Beyond

MPAI – Moving Picture, Audio and Data Coding by Artificial Intelligence is the first body developing data coding standards that have Artificial Intelligence (AI) as its core technology. MPAI believes that universally accessible standards for AI-based data coding can have the same positive effects on AI as standards had on digital media.
Elementary components of MPAI standards – AI Modules (AIM) – expose standard interfaces for operations in a standard AI Framework (AIF). As their performance may depend on the technologies used, MPAI expects that competing developers providing AIMs will promote horizontal markets of AI solutions that build on and further promote AI innovation.
Finally, the MPAI Framework Licenses provide guidelines to IPR holders facilitating the availability of compatible licenses to standard users.

Close

2019

A. Artusi, F. Banterle, F. Carrara and A. Moreo

Efficient Evaluation of Image Quality via Deep-Learning Approximation of Perceptual Metrics

Image metrics based on Human Visual System (HVS) play a remarkable role in the evaluation of complex image processing algorithms. However, mimicking the HVS is known to be complex and computationally expensive (both in terms of time and memory), and its usage is thus limited to a few applications and to small input data. All of this makes such metrics not fully attractive in real-world scenarios. To address these issues, we propose Deep Image Quality Metric (DIQM), a deep-learning approach to learn the global image quality feature (mean-opinion-score). DIQM can emulate existing visual metrics efficiently, reducing the computational costs by more than an order of magnitude with respect to existing implementations.

Close

@ARTICLE{8861304, author={A. {Artusi} and F. {Banterle} and F. {Carra} and A. {Moreno}}, journal={IEEE Transactions on Image Processing}, title={Efficient Evaluation of Image Quality via Deep-Learning Approximation of Perceptual Metrics}, year={2020}, volume={29}, number={}, pages={1843-1855}, abstract={Image metrics based on Human Visual System (HVS) play a remarkable role in the evaluation of complex image processing algorithms. However, mimicking the HVS is known to be complex and computationally expensive (both in terms of time and memory), and its usage is thus limited to a few applications and to small input data. All of this makes such metrics not fully attractive in real-world scenarios. To address these issues, we propose Deep Image Quality Metric (DIQM), a deep-learning approach to learn the global image quality feature (mean-opinion-score). DIQM can emulate existing visual metrics efficiently, reducing the computational costs by more than an order of magnitude with respect to existing implementations.}, keywords={approximation theory;computer vision;feature extraction;image enhancement;learning (artificial intelligence);neural nets;deep-learning approximation;perceptual metrics;human visual system;HVS;image processing algorithms;Deep Image Quality Metric;DIQM;deep-learning approach;visual metrics;image quality feature;Measurement;Image quality;Visualization;Distortion;Indexes;Feature extraction;Convolutional neural networks (CNNs);objective metrics;image evaluation;human visual system;JPEG-XT;and HDR imaging}, doi={10.1109/TIP.2019.2944079}, ISSN={1941-0042}, month={9},}

Close