I'm curious about the advances in the area of image super-resolution (SR) that have given the best results to date, both perceptually (visually pleasing) and objectively (e.g. PSNR, SSIM criterias). I'm mostly talking about Single Image SR (SISR), but I'm open to input on Multiple Image SR as well.

From what I've researched, Photo-Realistic Single Image Super-Resolution Using a Generative Adversarial Network looks like the one, at least as far as SISR goes. However, I am unsure as to what alternatives are out there, that come close or maybe even surpass it (both in the field of neural networks and other, more traditional fields).

If this reaches anyone's interest but my own, perhaps this can be a question that gets new answers as time and advances move forward.

  • $\begingroup$ Could you please review my answer? $\endgroup$
    – Royi
    Mar 8, 2022 at 18:31

1 Answer 1


In our days the Deep Neural Network methods certainly are generating best results.
Due to the intense research going on in this field the best method is a moving target hence one can not pin point to one.

One generation before them the best methods were based on Dictionary Learning.
For example you can use the K-SVD for Single Image Super Resolution.

Those approaches / methods works pretty well and are even improved using the EPLL approach.

You should be aware those methods are really slow for real world images.
Probably you should invest in the DNN world.
Pick any one of the latest on ArXiv Sanity.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.