A common frustration for content producers is to have images that are too small for the page layout. For example, sometimes the only image they have to work with is a rasterized (pixel-based) image that’s 300px wide, but they need the image to be 600px wide. Their only choice is to either resize the image and accept that it will appear fuzzier and pixelated, or use a different image.
Pixelmator Pro has released a featured called ML Super Resolution that can make a raster image larger while maintaining its sharpness and detail. They’re using Apple’s Core ML 3 and neural network capabilities to achieve what was once thought impossible.
How Pixelmator Pro uses Machine Learning (ML) to upscale images
The Pixelmator Team published a detailed blog post about how they’re using ML to resize raster images while maintaining visual clarity.
It takes into account the actual content of every image, attempting to recognize edges, patterns, and textures, recreating detail based on our dataset and extensive training. When upscaling, it can make much better predictions because a red pixel next to a blue pixel can be a completely different type of texture or edge in different images even though, to the primitive approaches, they’re always the same.
The post included several real-life examples of how well the new upscaling feature works. They also made the original and comparison images available to download. I tested the ML Super Resolution feature on one of their smaller original images, and it was impressive to see how well it worked.
The new raster image upscaling feature is available now. Pixelmator Pro retails for $39.99 and is available on the Mac App Store.