Adobe has announced the new AI-based feature called Enhance Details, that can improve image quality up to 30 percent. The new feature will become available in Camera Raw, Lightroom Classic and Lightroom CC and it will introduce a better way to resolve fine details in images.
Enhance Details is based on convolutional deep learning models trained on billions of photos.
What Enhance Details does is called demosaicing. Demosaic is, in fact, a known procedure used when a digital photo is created from raw camera data. The composite red, green, and blue value of every pixel in the images that we see as digital photo are created by a demosaic process.
In the past, people have developed and used a wide spectrum of different demosaic algorithms ranging from very simple ones (averaging neighboring pixel values) to quite complex and advanced algorithms.
However, these algorithms can be very computationally expensive. Moreover, very often they introduce problems such as unnatural shifts in colors, losing details, blurring etc.
The new feature – Enhance Details is, in fact, a demosaicing algorithm based on a Convolutional Neural Network. The network was trained on over a billion problematic samples optimizing for maximum image quality. This means that each image had one or more of the major issues that cause demosaic algorithms to fail.
Adobe reports that Enhance Details can give up to 30% higher resolution. An external GPU is needed to support the necessary computations for Enhance Details to work. More about Enhance Details can be read at the official blog of Adobe.