With the continuous advancement and adoption of emerging technologies, it appears that no industry or field will be spared. These technologies including artificial intelligence and machine learning have proven to have the potential to make things easier and efficient. For instance, only two approaches existed for taking usable pictures, particularly in low-light conditions. They include using a slow shutter, which calls for a tripod to get rid of blur or boosting the sensitivity of your camera’s sensor electronically, which brings in undesirable noise artifacts. Thanks to machine learning, a third technique has been found.
The new approach utilizes machine learning to artificially enhance the brightness of a dark photo after taking it in low-light conditions. The results of this method have been nothing short of stunning. Researchers at the University of Illinois Urbana-Champaign and Intel are behind the creation of this method, which may be the ideal post-production tool for professional photographers.
The tool will be most suitable for those experts who shoot in low-light environments such as capturing nocturnal wildlife at night or taking pictures of performances at concert venues. It can also be utilized in improving the quality of pictures taken using smartphones, particularly in dark bars.
Similar to the other numerous image processing creations nowadays, the research behind the latest tool was published in a paper dubbed “Learning to See in the Dark.” According to the paper, this new approach works by leveraging deep learning techniques in training an algorithm how a poorly exposed image ought to be brightened and color-corrected appropriately, especially during post-production processing.
The researchers from Intel and the University of Illinois delivered a neural network backed by a dataset made up of 5,094 excessively dark short-exposure pictures and a similar number of long-exposure photos that displayed what the scene ought to look like with adequate exposure and lighting. What’s more, all the images involved had been taken with Fujifilm X-T2 and Sonya7S II cameras.
For a person who has gone through the hassles of using Photoshop, for example, in trying to correct photos taken in overly dark conditions, he or she will agree that the outcomes of the new algorithm are quite impressive. The tool transforms the images from what would have originally gone straight into the computer’s recycle bin to photos that are usable to a certain extent.
Although the tool ’s processed pictures are still not as good looking as they would have been if the similar scene had been shot with a tripod and long exposure, they are good enough in case you just want to take a couple of shots on a night out with your buddies. Even so, the approach relieves the burden of having to drag the extra gear everywhere you go to get pictures at night.
It may take long before the tiny sensors installed in smartphone cameras have the same potential possessed by the giant sensors in DSLR cameras. Nevertheless, with this groundbreaking algorithm operating in the background of a smartphone camera application, the smartphone sensors may not have to change.
Source Gizmodo
