To be more specific, real-time rendering.

Most of the academia I checked are engaged in mathematical and visualisation work, but very few seem to be doing any sort of research in real-time graphics. Also noticed that majority of the recent publications have come from the industry. Did this kind of situation exist before as well, just went unnoticed?

when you say "real-time rendering" do you just me doing it quickly?

I have no idea about the current state of real time rendering research but I have a few thoughts on the basis that it isn't a big field of research.

Engine developers are leading the way.

In my experience academic's are not good at writing production quality code and most of the speed ups are hacks or tricks which reduce quality / quantity in a way that the eye doesn't notice easily.

Its also easily parallelized with a gpu. Throw more cores at it, it will get faster.

Who wants to pay for this research?

I'm trying to think of another field which requires real-time rendering which looks good and is currently lacking. Even naval simulators that the Royal Navy use have crap graphics.

Its easy to do crap graphics quickly, who defines what is "good"? Its easier to go for perfection (reproduce reality) and let hardware handle the rest.

Because real-time rendering depends, for the most part, on approximating offline solutions. So "academic research in real-time rendering" would be mostly limited to implementation (which is what the industry does) rather than the underlying mathematics and physics.

To the contrary, I've seen an awesome amount of research from academic sources rather than game engine developers. I think I initially experienced the same as you have, but I believe as you continue looking about for recent research you shall find there's a great deal of it that is indeed academic.

You guys know that there's a whole conference dedicated to real-time rendering, right? Go ahead and take a look through the papers from this year or last year and see how many come from academia. I would also suggest reading through the history of PRT research, or Ravi Ramamoorthi's long history of frequency space rendering if you want some examples of research that goes beyond just "implementations". Another good place to look is any of the many recent papers focused on using spherical gaussians and other SRBF's for representing BRDF's, NDF's, and visibility functions.

Besides...it's not like any research that isn't specifically focused on real-time rendering means that it's only exclusively relevant to offline rendering. Real-time rendering often makes use of the same techniques, principles, and insights as offline rendering. This is especially true if you factor in things that are pre-computed offline for use in real-time applications or games.

Most academics are interested in developing new techniques, not in making them fast enough for real-time, nor in making existing off-line techniques fast enough for real-time.

It tends to follow the old "Make it work, then make it fast" mantra -- academia makes it work, industry makes it work fast. This is not true of every case, but I'd say its been the general rule, rather than the exception, for at least as long as there's been consumer applications of the work.

Yeah, I was thinking of Ravi Ramamoorthi myself. Berkeley graphics has a fairly good amount of significant real-time research, though I wouldn't necessarily name them the most frequent source of new real-time research; not at all! There's plenty of new real-time rendering research coming from all sorts of universities. You just need to search more.

I too have an interest in real-time rendering, but I had no idea that this was the nature of academic research vs industry research. I'm applying to become a student volunteer at this year's SIGGRAPH hoping to get more insight in the field. I guess most if not all the research will be geared towards offline techniques? It will be awesome nonetheless! *EDIT* I take that back, SIGGRAPH has featured quite a bit of real-time work.