One of the most annoying things to happen to me is when I find out that the image I snapped on my phone camera turns out to be blurred or pixelated when I want to print it. Granted, I don’t always print out my photos. But on the rare occasion that I do, and it turns out to be blurred, it really does bug me. It’s not something you can fix with photoshop either. It’s just blurred.
Well, thanks to a bunch of scientists and researchers at the Max Planck Institute for Intelligent Systems in Germany have found a way to create high-resolution versions of low-grade photos such as the ones you snap, by means of AI or artificial intelligence. The method is a new one to the standard single-image super-resolution (SISR) technology that is currently available.
A standard SISR software will attempt to perform a pixel-perfect reconstruction by adding extra pixels and averaging them with surrounding pixels on enlarged images. Alas, this leads to blurry images which also lack sharpness.
How does EnhanceNet-PAT work?
This is where EnhanceNet-PAT comes into play. Developed by the Max Planck Institute for Intelligent Systems, the method uses machine learning. The algorithm is given millions of low-resolution images. It would then scale these images up. It essentially imagines a higher resolution image and then proceeds to add pixels to the original low-resolution image accordingly. The algorithm learns from its mistakes and also learns the difference between images to increase accuracy.
They say practice makes perfect and that’s the same for EnhanceNet-PAT as well. Once successfully trained, EnhanceNet-PAT doesn’t need to be given original photos either. once the software is trained it no longer needs to be fed original photos.
With tools like EnhanceNET-PAT and even Adobe’s new lineup of AI-related tools for image retouching and manipulation, we can finally have all our favourite photos crystal clear and ready for print. Indeed an exciting time to be alive and living in.