JPEG wants to use AI for the next compression standard

The Joint Photographic Experts Group, the committee that manages the various JPEG image compression standards, has begun to explore the possibility of exploiting Artificial Intelligence technologies and techniques to create a new compression standard.

In a recent meeting held in Sydney, the group formally started work to identify possible AI-based methods that can serve as a development base for a new image compression codec. This is an initiative already launched last year, under the name of JPEG AI, which sees a special group at work that has the task of studying image codecs based on neural networks.

In the context of this initiative, contacts have already been made with the IEEE – Institute of Electrical and Electronics Engineers. To collect publications within the ” Learning-based Image Coding Challenge ” and, which will be presented at the International Conference of Image Processing scheduled in Abu Dhabi for the month of October.

The aim of the committee is to identify cases that can demonstrate that Artificial Intelligence is able to obtain better results in compression by working with large image databases. In other words, the hope is that by using Artificial Intelligence and training it on large image databases it is possible to identify image coding methods that can show better compression efficiency than traditional approaches.

This is not a completely new approach: in 2017, Netflix developed a codec based on Artificial Intelligence techniques and technologies to find the best meeting point between media quality and data efficiency, using it in emerging markets such as India where the public enjoys massively content on smartphones using mobile networks.