Source: Hiver

Google, the search giant, is developing a machine learning method to improve picture details and reduce network burden by almost a third. It is exciting to mock image-processing impossibilities where CSI investigators or “Blade Runner” zoom into the images to see more detail than a photo could possibly have recorded.

Google will allow you to see your pictures even more clearly

Now, the ads giant allows that kind of photo enhancement to your picture as well as keeps you from using your entire mobile phone data plan. It would not assist law enforcement in seeing the perp reflected in the retina of some person but it will make pictures of Google+ load even quicker and decrease network data usage by about 1/3rd.

That is a welcoming change for people as our requirements for gigabytes are increasing whereas our phone data plans are still very expensive. The technology of the search giant is called RAISR, which is the short form of “Rapid and Accurate Image Super Resolution.” It is an intelligent use of AI (Artificial Intelligence) that foretells how a specific picture can be magnified based on the actual transformation of numerous other pictures.

In addition to this, the RAISR has the ability to expand pictures without pixelated or jagged edges. Users, with a software like Photoshop, shrink a picture so that it has even fewer pixels. That process is called downsampling. This process is quite useful if you want to show a large original picture on a small screen, like on a smartphone.

Google’s approach for providing good quality pictures

There is also another process called upsampling that just adds more pixels to the picture but it not does improve the quality of the photo which is why it is not recommended. The ads giant has taken a different approach, neither it is like upsampling nor it is like downsampling, but it is processing a single picture.

What the approach of the ads giant does is compare pairs of high-resolution actual-world pictures with their low resolution and downsampled equivalents. The AI software of Google, with 10,000 such pairs, gets a good notion of how a specific detail in the low-resolution is corresponding to what was in the high-resolution equivalent of that photo.

Traditional upsampling techniques use the similar mathematical change to the entire photo, however, the approach of the ads giant applies a different magnification approach for every little patch of the picture. Now in its Google+ app, the search giant is using the RAISR approach for some Android smartphones. On Wednesday, John Nack, Google product manager, said in his blog post that using RAISR cuts the data usage of the app by about a third.

​Nack added, “In the coming weeks we plan to roll this technology out more broadly.” In 2016, the search giant had detailed RAISR tech in a research paper. In a blog post about the said tech, RAISR researcher Peyman Milanfar wrote that even though the ads giant is not the only organization working on smarter upsampling technology, RAISR is around 10 to 100 times quicker than other alternatives which lets it run on a typical mobile device in real-time.

 

Comments

comments