Algorithm that Corrects Multiple Flaws of an Image by Using Artificial Neural Networks

Algorithm that Corrects Multiple Flaws of an Image by Using Artificial Neural Networks

strawberry-2293337_960_720

Digital images are playing an important role in communicating information to humans. For instance, medical scans are life saving and photo camera snapshots preserve wonderful memories. However, digital images have many flaws. They are subjected to blurriness, missing pixels, grainy noise, and color corruption. Computer scientists from the University of Maryland have designed an algorithm which makes use of artificial neural networks so as to apply wide range of fixes to corrupted digital images. The algorithm works in a way that it can recognise what an ideal image should look like without any of the flaws mentioned above such as blurriness or missing pixels etc. In fact this algorithm has been designed to address multiple flaws in an image.

This research team also included members from the University of Bern in Switzerland and they tested algorithm by introducing degradations in high-quality images which were uncorrupted. They then used this algorithm in order to check if it repairs the damage. The test results showed that the algorithm almost returned the images to their original state. While traditionally different tools have been designed to address different problems of an image separately, this new algorithm can address different problems of an image at the same time.

Artificial neural networks algorithms are based on AI and are inspired by the structure of human brain giving them the ability to assemble patterns of behavior depending upon the data which is inputted much in the same process as the human brain learns new information. The human brain can learn a new language through repetitive exposure to sentences and words in specific contexts. Similarly, this algorithm is exposed to a large database of high quality and uncorrupted images. The algorithm has the capability to take in a humongous amount of data and extrapolate the complex parameters which define the images and variations such as text, light, colour, edges, shadows. Thus, it is able to predict what an ideal or an uncorrupted image would look like and can, therefore, fix the deviations or flaws from these ideal parameters in a new image.