Canon Unveils AI Image Processor That Deftly Corrects Mistakes In Digital Photos

There’s seemingly nothing artificial intelligence can’t do—from answering questions to offering companionship and generating artwork, it doesn’t seem as if the technology will ever reach the full extent of its capabilities.

Now, to add to the growing list, Canon has introduced its own neural network, based on a machine-learning image processor that aims to significantly improve any amateur photographer’s images.

According to PetaPixel, the tool hones in on three areas where major mistakes are often found in photographs—noise, color, and lens imperfections such as aberrations or blur. Using “cutting-edge technology,” the system automatically corrects images to make them better.

With the capabilities of its deep-learning algorithm, which has been trained on an extensive database of images in both RAW and JPEG throughout the development of cameras and lenses, Canon claims its system has expertise in “all imaginable subjects.” 

The company said its latest software differs from the usual noise reduction technology on the market, which often removes details or deteriorates image quality while getting rid of unwanted noise in the background.

Instead, its AI generator helps photographs remain clear and high-definition even after removing noise, making it now possible to produce clearer pictures or smoother skin textures while still amplifying the amount of light in a picture.

Furthermore, current image sensors detect only one of three primary colors—red, green, and blue—and fill in the rest using information from the surrounding pixels, which sometimes results in false colors occurring.

Its deep-learning technology will no longer face this issue, having been equipped with the Neural Network Demonsiac that helps to suppress wrongful color interpolation on subjects, such as jagged lines, moire, or false colors on clothing.

If that wasn’t useful enough, the system can aid photographers in correcting problems that occur due to lens issues with its Neural Network lens Optimizer, which helps solve image degradations, d improve a photograph’s resolution, or prevent blurring at the edges of a frame.

It’s unclear if Canon will incorporate this technology into its forthcoming cameras to automatically adjust pictures, or if it will release software for post-editing purposes. Nonetheless, the system certainly seems promising and could help shutterbugs worldwide capture better shots.

[via PetaPixel and Australian Photography, images via Canon]