RSS
 

Posts Tagged ‘upsampling’

AI-based upsampling tech creates high-res versions of low-res images

01 Nov

Upsampling image and video files usually leads to pixelation and soft textures, simply because algorithms are not capable of replacing non-existing image detail. But scientists at the Max-Planck Institute for Intelligent Systems in Germany have come up with a clever solution that is capable of producing better results than anything we’ve seen so far.

The team has developed a tool called EnhanceNet-PAT, which is capable of creating high-definition versions of low-resolution images, using artificial intelligence. It’s not the first attempt at solving the super-resolution task but the approach is a new one.

Talking to Digital Trends, Mehdi M.S. Sajjadi, a member of the research team said

“Before this work, even the state of the art has been producing very blurry images, especially at textured regions. The reason for this is that they asked their neural networks the impossible – to reconstruct the original image with pixel-perfect accuracy. Since this is impossible, the neural networks produce blurry results. We take a different approach by instead asking the neural network to produce realistic textures. To do this, the neural network takes a look at the whole image, detects regions, and uses this semantic information to produce realistic textures and sharper images.”

To train the technology, large image libraries were fed into the AI-based system to accumulate knowledge of textures. The neural network is then given the task of upsampling previously downsized images. Finally, the research team compares the results to the original high-resolution image and modifies the algorithm to correct for errors or inconsistencies—this way, the algorithm keeps improving and after some time can work without any human help.

According to the team, photographic applications for the network include upsampling old movies to 4K resolution or restoring old photographs. However, the system could also be used to enhance object detection, for example in image search or mobility applications, such as self-driving cars. It seems that cheesy “Enhance!” trope from CSI is about to become reality…

To learn more about this technology, check out the full research paper here.

Articles: Digital Photography Review (dpreview.com)

 
Comments Off on AI-based upsampling tech creates high-res versions of low-res images

Posted in Uncategorized

 

Google brings RAISR smart image upsampling to Android devices

14 Jan

Google first showed off its RAISR technology, which uses machine learning to produce high-quality versions of low-resolution images, in November last year. Now the company has published a blog post to announce that RAISR has been implemented into Google+ for Android. Google+ is used by many photographers to display high-resolution images and the move is aimed at reducing mobile data requirements, which could be particularly useful in areas with slow connections or when using data is expensive, for example when roaming. 

RAISR allows for viewing images at their (almost) full glory while reducing bandwidth requirements per image by up to 75%. Google has only just begun to apply the technology to high-resolution images in the Google+ streams of a subset of Android devices but is already processing 1 billion images per week, resulting in a total bandwidth reduction of about a third for the affected users. Google says it is planning to roll out RAISR more broadly in the coming weeks, so your data consumption might go down soon if you use Google+ frequently. 

Articles: Digital Photography Review (dpreview.com)

 
Comments Off on Google brings RAISR smart image upsampling to Android devices

Posted in Uncategorized

 

Google RAISR uses machine learning for smarter upsampling

16 Nov
 Top: Original, Bottom: RAISR super-resolved 2x, Image: Google

Upsampling techniques to create larger versions of low-resolution images have been around for a long time – at least as long as TV detectives have been asking computers to ‘enhance’ images. Common linear methods fill in new pixels using simple and fixed combinations of nearby existing pixel values, but fail to increase image detail. The engineers at Google’s research lab have now created a new way of upsampling images that achieves noticeably better results than the previously existing methods.

RAISR (Rapid and Accurate Image Super-Resolution) uses machine learning to train an algorithm using pairs of images, one low-resolution, the other with a high pixel count. RAISR creates filters that can recreate image detail that is comparable to the original, when applied to each pixel of a low-resolution image. Filters are trained according to edge features that are found in specific small areas of images, including edge direction, edge strength and how directional the egde is. The training process with a database of 10000 image pairs takes approximately an hour. 

Once RAISR has been trained it is capable of selecting the most appropriate filter for each pixel in a given low-resolution image to fill in new pixels in order to create a higher-resolution version. RAISR can also remove aliasing artifacts, such as moiré patterns or jagged lines in the low-resolution images when creating the larger version, something that linear methods are not capable of doing.

Eventually, Google may be able to use the technology to upsample images that are sent at mobile bandwidth-friendly resolutions. More information on RAISR is available on the Google Research Blog.

 Left: Original, Right: RAISR super-resolved 3x, Image. Google

Articles: Digital Photography Review (dpreview.com)

 
Comments Off on Google RAISR uses machine learning for smarter upsampling

Posted in Uncategorized