Guest_Jim_* Posted November 9, 2015 Posted November 9, 2015 The human eye is pretty good at finding deviations and variations, but computers struggle to do so, unless the differences are on a large scale. By giving computers the power to spot these differences though, they could help us identify defects and create better looking graphics. Now researchers at MIT have developed a pair of algorithms that are capable of doing just that from still images. One of the algorithms works by finding repeated shapes in an image and then creates an ideally regular, target version of the image. Next it moves around the pixels of the original image, to approximate the target version, and repeats the earlier process to create a new target image. It continues to iterate these steps until the target images look the same, leaving a surprisingly natural looking image. Reversing this process will amplifying deviations in the original image. The other algorithm looks for color differences to identify shapes, generates idealized versions of these shapes, and then exaggerates the deviations between them and the image. Potentially these algorithms could be used to identify structural defects, camouflaged objects, and movements our eyes would not normally catch. They can also be applied to create more polished graphics with image-editing software. Source: MIT Back to original news post Share this post Link to post Share on other sites More sharing options...
Recommended Posts