Feb. 27, 2023 - Some research institutions expressed concerns like: “We don’t want that too many burdens are put on our algorithmic developments. Advanced solutions have been rejected due to complexity constraints, but a few years later these constraints were not an issue anymore.”
I can only say that this was absolutely true in the past. However, if the complexity is growing much faster than the technology capabilities, the new standard will be in trouble. Either you have to wait many years till the technology is available to implement the full quality of the standard and the time to market becomes far too long to lead to a commercial success, or you compromise the quality to the point the improvement is not good enough to justify the RD investment required by the new standard.
In other words, feasibility with available technology is key to success. If we look at VVC we are now 3 years after the standard was published and there is no real time encoder available capable of delivering the full quality promise of the standard. We don’t expect this to happen at best before 2025, five years after the standard was issued.
This situation should alert us because it is the first time it happens. For instance, HEVC was much faster and full quality real time encoders were on the market two years after (2015 versus 2013).
How Do These Challenges and Objectives Impact the Planet?
Environmental sustainability is also a new strong requirement. Extra complexity means extra power consumption. Each standard requires more power despite the great progress of the underlying technology.
Future Concerns – Will Machine Algorithms Replace Human Workers?
Does it mean that researchers in image coding are going to run out of work? Absolutely not. The direction and objectives of the research for the next coming years should be in priority to reduce the complexity (or at least keep it growing slower than the underlying technology) and not to first improving the compression rate like in the past. There is a lot of great research actions to address for many years to achieve this goal.
Challenges Facing Hardware vs. Software in 2023
“These are mainly addressing hardware implementations. However, our focus is more on the software side and there are many applications which are mainly based on software solutions.”
I think that it is not a matter of software or hardware. Complexity is hurting both SW and HW. Feasibility in 2023 of a real time, full quality, encoder with VVC for instance is the same challenge. In addition, SW is worse than HW in terms of power consumption. There is an order to magnitude in favor of HW.
Therefore all my comments are applying even more to SW. Never forget that you need HW to run a SW codec and generic CPU/GPU are using much more power than optimized HW implementations.
Visit VITEC GreenPEG initiative: https://www.vitec.com/blog/editorials/vitec-launches-greenpeg-initiative-to-optimize-clean-environmentally-friendly-production-standards-to-reduce-greenhouse-gas-emissions/ for more information about impactful sustainability measures.
See the original post for details: https://www.vitec.com/blog/editorials/adaptive-challenges-and-objectives-for-a-new-video-compression-standards/
by Philippe Wetzel, CEO and Founder of VITEC