Artificial intelligence technology is advancing very fast and now a tool has come up that clears very blurred images of faces.
Scientists at Duke University in the United States have released the results of a study that found that the new algorithm converts blurred or pixel images into real-looking faces.
According to researchers, the quality of images is excellent, but does not work on fine images, but the best results can be seen only on blurred images.
This algorithm is called Photo Up Sampling Latent Space Exploration, which is currently in the concept phase, but can be tested online by visiting this link.
The researchers said that never before had super-resolution images been created with such detail, which makes the image 64 times better than the original resolution.
He added that the tool was not designed to identify people and that it could not turn a CCTV’s blurred face into a high-resolution image of a real person, as the faces may look real but made by the tool itself. Which may be different from the original.
The researchers said that this tool would help in correcting any poor quality image.
Photo courtesy Duke University
Earlier, in such tools, low resolution image blur was removed, but the result was not so good.
To avoid this problem, Duke’s researchers incorporated a number of high-resolution faces developed by AI into the tool to improve the quality of the faces as much as possible while working on the original image.
The researchers’ goal was to create a high-resolution image from a low-resolution image, and they seem to have succeeded.
The research team said they were able to convert a 16×16 pixel image to 1024×1024 in a matter of seconds.
He said that this technique is not limited to faces only, but it can be used to turn a low resolution image of anything into the best image that looks real.