Google RAISR Makes Low-Res Images High Quality
One of the biggest issues for smartphone users is bandwidth usage. There are very few true unlimited data plans anymore. This has become a problem due to the rising sizes of web pages. Most web pages now average 16MB and are growing rapidly every year. Ad blockers help on the desktop but there is no easy solution on smartphones. Google just made a big breakthrough in viewing images online.
Google scientists have used machine learning to keep the sizes of image small without any corresponding reduction in quality. According to the scientists, the resulting file size is 75% smaller than the original image.
This new technique is RAISR. It stands for “Rapid and Accurate Image Super-Resolution.” Normally, when resizing an image, it loses quality. RAISR can take a low-quality image and increase its resolution. It does this while maintaining detail during the upsampling process. It enhances the end image by using filters.
Almost anyone who has used an image editing application knows what happens when you take a smaller image and try to make it bigger. The result is a blurry and sometimes blocky image. When comparing images processed with RAISR, the gain in quality is clear.
The RAISR system has been trained using machine learning. The training involved feeding low-quality and high-quality version of the same image. This way, it was able to figure out which filters worked best to restore the most amount of detail.
