Scientists at NCFU Speed Up JPEG XS Image Transmission Without Sacrificing Quality
Scientists at North-Caucasus Federal University have adapted an earlier method for accelerated image processing to the modern JPEG XS coding standard, cutting data transmission time in half without sacrificing image quality.
Imagine a surgeon waiting for an MRI scan, a satellite transmitting gigabytes of data about a forest fire, or a factory production line where even a millisecond of downtime matters. In each case, one requirement takes priority – getting the image quickly without sacrificing quality. That is the challenge scientists at North-Caucasus Federal University set out to address by adapting a parallel image-processing method to the international JPEG XS standard.
The idea behind the technology is easy to understand, even if implementing it is not. Instead of processing each pixel individually, the algorithm handles entire groups – from two to five pixels at a time. The results are striking: compression performance increased by 109%, while image reconstruction performance rose by 144%. In effect, processing and data transmission time was cut in half.
Pros and Cons
The technology is not intended to make photos load faster in a messaging app. Its target applications are systems where latency is measured in fractions of a frame and image quality is critical to decision-making. These include medical imaging, satellite-image analysis, industrial inspection, environmental monitoring and forecasting hazardous events.
For Russia’s IT sector, this is not yet a finished commercial product but an algorithmic foundation – a basis for future hardware accelerators, codecs and computer vision systems. For the public, the benefits would be indirect but tangible: faster diagnostics, more timely environmental data and more accurate satellite monitoring.
The technology’s weak point is higher energy consumption. During encoding, power use increases by about 20 times; during image reconstruction, it rises by nearly 125 times. That means the technology is designed for specialized applications where electricity costs take a back seat to latency requirements. At this stage, a mass-market consumer product is still out of the question.
Prospects: From the Lab to Hardware
In the near term, the most realistic path is to deploy the algorithm within Russia. The practical groundwork is already taking shape. In 2024, Synterra Media and Stream Labs tested Russian-made Agat equipment using JPEG XS for professional television broadcasting. In 2025, the FPGA-based MediaXGate platform was unveiled with support for the same standard in high-load broadcast-center nodes. Potential Russian customers for NCFU’s technology already exist.
The technology could also have export potential: the algorithm is compatible with the current international standard, whose third edition was released in 2024. However, publicly available sources so far contain no information about an industrial prototype, foreign partners or commercial trials. Reaching the market will require reducing energy consumption, implementing the algorithm on FPGAs or specialized processors, and testing it on real-world data.
A Five-Year Journey
The latest work did not come out of nowhere. Since 2021, NCFU has been researching high-performance devices for medical image processing. In 2022, the university received support for projects to develop hardware accelerators for three-dimensional visualization. In 2023, an earlier version of the technology based on the Winograd algorithm and wavelet transforms was presented. The latest result is a logical continuation of that work, adapted specifically to JPEG XS.
A Scientific Result, Not a Market Breakthrough
NCFU’s technology is a significant scientific result and part of Russia’s growing domestic expertise in high-performance computing. Its strength is a substantial speed increase while remaining compatible with an international standard. Its main limitation is energy consumption. The most likely scenario over the next few years is the development of a specialized accelerator for applications where latency matters more than computing costs. Until a working product emerges and independent testing is conducted, the development should be viewed as a scientific achievement rather than a technological breakthrough for the mass market. But the groundwork for such a breakthrough is already in place.
