Digital image compression: after JPEG

The JPEG compression format is more than 25-years-old: a canonical age in the digital age. A UCL researcher and spin-off have imagined the compression standard of tomorrow: JPEG XS.

For 25 years, the JPEG image compression format has become widespread across the planet. This digital storage format makes it possible to compress the images via a standardised algorithm that can be shared among all users. The goal is to reduce the space occupied on hard disks and internet bandwidth.

‘In general, image and video compression is a research topic that has been investigated for a long time’, explains Antonin Descampe, a researcher and lecturer at UCL’s Institute of Information and Communication Technologies, Electronics and Applied Mathematics  (ICTEAM) and co-founder of the IntoPIX university spin-off. ‘One might think it’s no longer necessary to compress digital documents further, since networks are now larger and hard disks more spacious. However, the increase in file size goes hand in hand with these developments. Whereas before, a DVD contained images with a resolution of less than 1,000 by 1,000 pixels, today we’ve moved to high definition (HD, 1920 x 1080 pixels). Every day we see more "4K" televisions (4000 x 2000 pixels), and we are now talking about "8K" (8000 x 4000 pixels).’

Finding the right compromise

Despite the improvement in sharing infrastructures, images and videos are increasingly in need of space. This evolution has created a clear need to invent a new compression standard. ‘The goal of the research,’ Dr Descampe says, ‘ was to find a compression standard that is both uncomplicated and that can pass through existing channels, and with "visually lossless" quality. Image compression always involves some loss of quality. But the purpose of these standards is to find a compromise between, on the one hand, compression efficiency, that is, how much can be compressed while ensuring a sufficient level of quality, and on the other hand, the complexity of the algorithm that directly impacts the cost of implementing this standard.’ Indeed, this algorithm must then be concretely implemented in printed circuits. However, the more complex the algorithm, the more expensive the circuits.

The right compromise, ‘intoPIX’, a UCL spin-off, found it three years ago. ‘UCL has always been active on standardisation committees’, Dr Descampe says. ‘But in this case, we’re leading the way.’ This young researcher is also the co-founder of the spin-off, created in 2006 on the basis of research carried out at PiLAB, directed by Benoît Macq, professor of electrical engineering at the Louvain School of Engineering. The staff of ‘intoPIX’ have worked on a low compression algorithm: an algorithm for lightweight image compression, without loss of visual quality and for use in existing channels to minimise implementation costs. This is the birth of JPEG XS. ‘When we implemented the JPEG XS algorithm, we were convinced that it could become a standard’, Dr Descampe says. ‘Three years after submitting it to the ISO committee, the standard’s quality assessment has just been completed. JPEG XS should become an international standard in early 2019.’

Many fields are interested in this standard, as the JPEG XS can carry four to ten times more information than if it was uncompressed on a given channel. The space sector is particularly interested: the possibility of transmitting superior quality information through existing channels and therefore with negligible costs is very promising. There’s also potential for video production of live events, virtual reality or augmented reality applications, the automotive industry (cars are equipped with a growing number of video sensors), camera manufacturers and cameras, etc.

Moreover, the standard is already used in ongoing research projects within PiLAB, particularly in the field of optimising the consumption of video compression circuits or in the field of miniature cameras for drones and private cars.

 

Interview

What process did you go through to validate the existence of this new standard internationally?

Contrary to what one might imagine, ‘JPEG’ is the name of a group of experts that meets under the auspices of the International Organization for Standardization (ISO). This ‘JPEG Committee’ has existed for 30 years and meets three or four times a year to work on changes in the standard, which has been a huge success: millions of images in JPEG format are exchanged today on the internet, which is proof of its success. But needs can evolve, applications change...The Internet of 25 years ago is not the one we find today. And in our case, we identified a need for a new standard of image compression that we ended up calling JPEG XS.

The validation process takes a long time. A particularly intense moment was the meeting we had in Chengdu, China, in October 2016. This is the meeting during which we had to choose the basis of the future JPEG XS standard from among six technology proposals. After three days of intense discussion of the evaluation results of the various proposals, what a relief and what pride to see the choice was made to focus on the technology proposed by intoPIX, the ‘small’ UCL spin-off that dared to swim with the big fish.

Will we soon use only JPEG XS?

No: the purpose of the JPEG XS standard is not to replace the JPEG. Replacing a standard that is so widely disseminated and integrated with all electronic devices around the world is neither feasible nor desirable. It is rather in specific areas that we want to intervene. Especially in the transition phase: rather than asking people to replace all their infrastructure because video goes from 4K to 8K and they can’t support it physically, why not propose a standard that compresses very slightly, without visual loss, and keep the current infrastructure? Replacing an entire network that supports HD in order to accept 4K is extremely expensive. JPEG XS should be seen as a facilitator, which will be used regularly during market changes.

 

A glance at Antonin Descampe's bio

Antonin Descampe was born in Brussels in 1979. He received his bachelor’s in engineering in 2003 and continued his higher education at UCL. In 2006, he co-founded the intoPIX spin-off with four other engineers: Gaël Rouvroy, Jean-François Nivart, François-Xavier Standaert and François-Olivier Devaux. The company specialises in image compression and security. In 2008, he obtained his PhD, focusing on hardware implementation of the JPEG 2000 standard and its use for searching images in large databases. Overall, his main research concern is image and video compression (JPEG 2000, JPEG XS, HEVC). In addition, he manages the OpenJPEG project, a leading open source software for JPEG 2000. As a guest lecturer in UCL’s Faculty of Economic, Social and Political Sciences and Communication (ESPO), Dr Descampe gives a course on the emergence of new technologies and their applications. ​​​​​​

Publié le 05 juin 2018