Skip to Main Content
March 13, 2024

IntoPIX Updates us on JPEG XS

Ben Schwarz wrote an interesting article on the 8K Association site about the JPEG XS codec which can be used for low latency but visually lossless compression of RAW and video data in production environments. Ben spoke to Jean-Baptiste Lorent, Sales & Marketing director of IntoPIX which is a key provider of the technology. We caught up with Lorent again to see what has changed in the two and a half years since his article.

Version 3 On the Way

One of the key changes that has happened to JPEG XS is that it is the process of finalizing and publishing Version 3 of the specification. This is a bit more complicated than you might think as there are five different sections and each is finalized separately. They should be approved in the next six months or so. A key part of Version 3 is the availability of TDC Profiles which really change things for the codec.

Before we dive into the details of TDC profiles, for those that are not close to the subject, what is JPEG XS? The codec is a very low latency, low footprint codec designed for content capture and production or applications that need extremely high (visually or mathematically lossless) compression. Unlike other codecs, it was developed ‘backwards’ Lorent told us.

Rather than starting from a desire to simply compress to essentially 50% of the previous generation, the JPEG XS codec started with a requirement for low complexity (few gates) and very low latency as well as visually lossless capability (based on ISO standards). The target was to allow compression of 4K60p content with an older Intel i7 CPU or with less than 50% of the power available in low cost FPGAs. Latency from end to end was set at max 32 lines.

The TDC Profile

The new TDC profiles use a concept called Temporal Differential Coding. That is to say, the codec takes account of changes over time as well as the spatial features of the image. Previously JPEG standards have focussed on spatial features as MPEG standards do. The codec exploits the concept of wavelet encoding but TDC introduces the idea of wavelet prediction. The codec does not use motion compensation to reduce the complexity of the processing. Lorent told us that the codec uses compression to create a frame buffer and keep it small to avoid external memory access which adds latency and takes a lot of power.

One of the features of the profile is that a basic frame can be sent with a (relatively) high level of compression, but if there is a static image, the subsequent frames can continue to add detail to get to a lossless image. Lorent gave the example of medical applications where you might need the low latency and high quality for applications such as endoscopy, but where static images are shown with a very high level of accuracy.

Note how in the PSNR measurement on the right hand chart the image quality ramps up sharply in the first half a dozen frames.

Despite the high performance of the TDC profile, the additional work must not add to the latency and Lorent showed that the profile only adds a handful of lines to the complete encode decode process, which can stay within 20 lines.

Applications 

Although it was designed from a broadcast and content creation perspective, JPEG XS also has wider applications. It can be used for digital signage and other ‘monitor like’ applications as it works well with graphic content as well as photographic or natural video material.

JPEG XS is a key component of IPMX, an open standards initiative for AV Pro applications and based on ST-2110 IP-based distribution. The IPMX has backing for groups including the VSF, EBU and SMPTE and is widely expected to be strongly adopted.

The codec could also be used for other low latency, high quality, screen-based applications such as remote KVMs. The writer was reminded of work going to replace control towers at remote airfields with high quality cameras providing human-vision level detail. As previously mentioned, remote medical applications are also very likely to be enabled by the standard.

JPEG XS is being developed to support cloud applications and is part of the AWS Cloud Digital Interface (CDI) for cloud to ground and ground to cloud use (although CDI is currently limited to 4K).

Other features

 One of the nice features of JPEG XS is that it is relatively simple to extract a lower resolution stream from the main one with no additional latency. This means that, for example, a single primary stream put into a 1Gigabit Ethernet switch can be used to create 2K or 1K streams. It also has applications in feeding multi-viewers and video switchers via the cloud or over ST-2110.

So what is the Difference between JPEG XS and TicoXS?

 Something that has been on the writer’s mind is the question of the difference between the IntoPIX TicoXS technology and JPEG XS (on which TicoXS is built). Lorent explained that as well as supporting the standard, TicoXS can also take advantage of proprietary features and technologies that either have not yet gone through a standardization process or not suitable for standardization. Standardization also takes a lot of time which can slow down the introduction of new ideas.

0 0 votes
Article Rating
Subscribe
Notify of
guest
0 Comments
Inline Feedbacks
View all comments
0
Would love your thoughts, please comment.x
()
x