As smartphones continue to shrink in size, it’s becoming increasingly difficult for manufacturers to upgrade their camera modules without adding pesky bumps on the back. Therefore, a slew of companies has begun to shift their focus to improving the pictures you take through software. Google last year proved how powerful this approach could be, but it’s not done just yet.
The search engine leader has partnered with a team of MIT scientists who have devised various new algorithms which are capable of retouching your pictures like a professional photographer. For achieving this level of accuracy, the researchers along with employing machine learning frameworks trained a bunch of neural networks on a set of 5,000 pictures. These images were created by Adobe and MIT, and each picture was manually altered by five different photographers. The algorithms took this data as an input and basically, learned what kind of specific improvements are required to be made on various photos.
For instance, if the photographer tweaked brightness and exposure of an underexposed picture, the algorithm will automatically do something similar if you capture an underexposed image. Of course, these calculations need to be optimized enough for being able to process on a smartphone. A vast number of companies still haven’t figured HDR processing which causes a lag of an extra second or two once you tap the shutter button.
Fortunately, the team is aware of this and mentions that the software itself is no bigger than a single digital image. In addition to that, these patterns can even be reproduced for representing an individual artist’s style and used as filters as well, something which Prisma does already. However, it’s currently unknown when this technology will make its way into mainstream smartphones. Perhaps, Google is already working on integrating these algorithms in the upcoming Pixel smartphones. We’ll know soon in the coming months what the company has in store for us.
This technology has the potential to be very useful for real-time image enhancement on mobile platforms Google researcher Jon Barron.
Using machine learning for computational photography is an exciting prospect but is limited by the severe computational and power constraints of mobile phones. This paper may provide us with a way to sidestep these issues and produce new, compelling, real-time photographic experiences without draining your battery or giving you a laggy viewfinder experience.