Patent reveals Nintendo is working on upscaling technology

Image: Nintendo

Patents are always a pretty intriguing thing in the tech industry – sometimes they represent ideas and products that will never see the light of day, but sometimes they give clues about future publications. A fresh one US patent application could be the latter as they are technologies that could allow Nintendo to improve the graphics in its hardware.

The application opened on March 25, 2020 and was released to the public yesterday (September 30). The application is entitled “Systems and Methods for Machine-Learned Image Conversion”. In NVIDIA’s case, this is short for Deep Learning Super Sampling, which works on some of its GPUs to increase image resolution and Real-time quality with remarkable efficiency and less load on the graphics card. It’s an impressive piece of technology, and it’s been the focus of many conversations about how Nintendo could create a new Switch-style wearable device that outputs higher-resolution graphics while operating at lower power. Digital Foundry has examined this in detail.

What’s fascinating about this application is that Nintendo is clearly researching this in-house – a named party to the application is Alexandre Delattre, co-founder of Nintendo European Research and Development. The “Introduction” of the patent also recognizes that this is an area that is being explored across the industry:

Machine learning can give computers the ability to “learn” a particular task without specifically programming the computer for that task. One type of machine learning system is called convolutional neural networks (CNNs) – a class of deep learning neural networks. Such networks (and other forms of machine learning) can help, for example, to automatically recognize whether a cat can be seen in a photo. Learning takes place using thousands or millions of photos to “train” the model to recognize when a cat is in a photo. While this can be a powerful tool, the resulting processing from using a trained model (and training the model) can still be computationally intensive when deployed in a real-time environment.

Image up-conversion is a technique that converts images created in a first resolution (e.g. 540p resolution or 960 × 540 with 0.5 megapixels) to a higher resolution (e.g. 1080p resolution, 1920 × 1080, with 2, 1 megapixel). This process can be used to display first resolution images on a higher resolution display. For example, a 540p image can be displayed on a 1080p television and can (depending on the type of upconversion process) be displayed with increased graphical accuracy compared to when the 540p image is directly upscaled using traditional (e.g. linear) upscaling on a 540 TV. Different image upconversion techniques can compromise between speed (e.g., how long it takes to convert a given image) and the quality of the upconverted image. For example, if an up-converting process is performed in real time (e.g., during a video game) then the image quality of the resulting up-converted image may suffer.

Accordingly, it goes without saying that there is a constant search for new and improved techniques, systems and processes in these technology areas.

Ultimately, it should come as no surprise that Nintendo is researching machine learning upscaling, as this will likely be a determining factor in the event the company chooses to keep a Switch-style form factor and offer greater graphical fidelity in the future. It is also interesting whether Nintendo will still use NVIDIA technology in future devices; If it is developing its own solution, it may not need NVIDIA’s DSL solution. Of course, depending on what and who you believe, there are reports that ‘4K’ development units are already in the wild.

Let us know what you think in the comments!

Source link


Leave A Reply