This might just be the first camera manufacturer to publicly make any serious attempts at implementing deep learning AI into their post processing. Sure, we’ve seen it from the likes of Adobe, Skylum, Topaz Labs and others, not to mention massive implementation in smartphones, but I don’t think camera manufacturers have made that big a deal of AI yet – except for autofocus subject tracking.
Canon, though, has shown off its new deep learning AI tech for post processing images, and the results look pretty impressive so far. It’s currently unknown when Canon might release this tech to the public. We also don’t know that when it is released, if it’ll just be available in Canon’s Digital Photo Professional for the desktop or Digital Photo Professional Express for the iPad, or if Canon will actually implement it into their cameras.
Canon’s goal with the software appears to focus mostly on removing noise, blur and moiré from photographs. Problems that can still exist no matter how good your equipment is. Sometimes the circumstances of the shot require us to shoot beyond the capabilities of that gear. Perhaps it’s just far too dark for us to handhold steadily, even if we ramp the ISO right up. Or perhaps the lens isn’t sharp from corner to corner.
The original announcement from Canon Japan is in Japanese, although Google does a good job of translating things into a way that’s at least understandable to give you an overview of the issues and how Canon is attempting to overcome them with its new deep learning AI tech.
[There are sometimes] unavoidable problems with the image quality of the photos. For example, noise that makes a photo look grainy, moiré that makes a moire pattern that should not exist, and blurring of an image caused by the principle of a lens. There were times when information was reflected. When a wide-angle lens is used, the image quality in the peripheral areas off the center of the lens tends to be blurred due to the deterioration of the optical performance of the lens.
As AI technology advances and deep learning technology is being applied in a variety of ways, Canon, as a leading company with extensive knowledge of cameras and lenses, is committed to the inevitability of the principle of photography. In order to solve this problem, he directly tackled the development of deep learning image processing technology himself.
Canon’s deep-learning image processing technology, which aims to achieve true-to-life photography, covers three areas: noise reduction, color interpolation, and aberration diffraction correction (lens blur correction).
As with all deep learning systems, the system uses a huge database of images that Canon says have been accumulated through the development of cameras and lenses so far, covering a wide range of subjects. These images are then fed into the system, allowing it to do its thing. As well as noise, moiré and lens aberrations, it also seems to be able to handle glow from lights and other high contrast scenes quite well, making subjects appear much more clearly in the final result.
As mentioned, Canon hasn’t yet said when (or even if) they’re going to turn this tech loose on their paying customers yet, although I think it’s only a matter of time. Whether or not it’ll just be confined to Canon’s desktop & iPad editing software remains to be seen. It might come to some cameras, although it’d probably only be useful when shooting jpg, not raw, but I wouldn’t hold your breath.
You can read Canon’s complete announcement and description of the technology on the Canon Japan website.