The VMAF metrics work by comparing renditions against a pristine reference video. However, this is not what the human mind does when viewing content. A metric that evaluates content for signs of coding-induced degradation without the use of a reference image would be desirable to have.
Have the project maintainers considered using/including such algorithms in VMAF? I imagine for Netflix scenario it is not relevant as there is always pristine reference available but I thought I would ask just for informational purposes.
The VMAF metrics work by comparing renditions against a pristine reference video. However, this is not what the human mind does when viewing content. A metric that evaluates content for signs of coding-induced degradation without the use of a reference image would be desirable to have.
Have the project maintainers considered using/including such algorithms in VMAF? I imagine for Netflix scenario it is not relevant as there is always pristine reference available but I thought I would ask just for informational purposes.