How do you upscale an image without loss of quality? Is that even possible? Technically if you just enlarge digital image files, there will be always a quality issue. Visual information does not just reproduce as you stretch a graphic. As part of our work here at TechAcute, we try to get the best possible images for our posts but sometimes there’s only one source file and in a worst-case scenario, it’s not very large. So is this something we have to accept? Or do you just need to pick another larger and better-looking image instead? I went to do a bit of research and here’s what I came up with.
Deep Image lets you upscale your images in your browser or smartphone
I found out that there are solutions out there that help you to resize your images and avoid getting any sort of quality issues or visual artifacts by not only increasing the resolution but also adding in image information to “fill the gaps” based on an algorithm. There might be various solutions around but I tested Deep Image, a neural network image enhancer. It works in your browser but you can also try out their app for Android or iOS.
Also interesting: TecoGAN – Increasing Media Resolution without Losing Details [Video]
Their primary function is upscaling images without noticeable loss of quality, upscaling video frames for animation, enhance images for ecommerce purposes, and enlarge images for print. The mobile app info pages on the respective app stores also advertise features such as repairing blurred selfies, upscale photos t high resolution, removing noise from photos, and enhancing pixelated images.
How does it work?
You just load up the Deep Image website, register or sign in with an existing account, select the image file on your hard drive that you want to process and select the sizing and correction options you need. It will then be queued up for processing and depending on how busy the service is, it might take a minute or two before you can get your optimized and upscaled image file. If you go for a premium account you don’t have to wait. That’s already the whole process. Even if you double the resolution of an image, the filesize should still remain more or less the same but there is no guarantee to that particular angle, as it depends heavily on what kind of image you might be processing. Please find an example image processing result below.
There is no magical tool that will fix all of your image processing problems but this is a curious step forward and even though this won’t make the impossible possible, it improved the possibilities of what you can do with image upscaling without blurring the image out or getting noise. Over time “AI” and neural networks that support these functions will become more sophisticated and the usefulness of such tools is likely to increase further in the future.
Free service and premium pricing overview
If you want to use the browser app only every now and then you’ll be okay with the free tier of the Deep Image service, but if you’re looking for a commercial solution, you can check out their pricing on the business service. You can make a purchase for a one-time payment like five images for $1.99 paid once, or a service subscription for 100 images for $9 on a monthly basis. If you need bigger volumes you can get cheaper plans but these are the two smallest examples for each payment option.
Other AI image upscaling solutions
Deep Image is maybe an interesting solution but it’s not the only one out there. If, for whatever reason, you don’t like Deep Image, you can also check out the following options:
Further reading
Do subjects like this one interest you? Here are some more posts that might interest you:
- How Nvidia VRSS Boosts Image Quality in VR [Video]
- Neural Talking-Head Video Synthesis Meets Video Conferencing [Video]
- Direct Comparison: TAA vs DLSS Render in 3DMark [Video]
- Pelican Imaging Adds Depth for Augmented Reality Devices
YouTube: ENHANCE! Upscaling Images with Neural Networks (Geoffrey Litt)
Photo credit: The feature image has been done by Cal Gao. The example image in the body of the article has been done by Braedon McLeod.
