Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Possible sRAW for DSMC?

Before anyone says, "can't you guys do it like XXXXX?", I'd suggest putting XXXXX through the same scrutiny that RED is subject to BEFORE asking the question. Why would we want to do sRAW when it sucks so bad?

There is an assumption with the "big guys" that they must be doing it right... when that is not necessarily the case.

Just saying...

Jim
 
Heh. You guys keep dissing sRAW, but I am not seeing it in the results I am shooting with the 5D2 in 1/2 res sRAW. In fact, the results are absolutely stellar (from what I can see). Granted, I only have a layman's eye for this type of thing.

One thing I wonder about is this: If you are shooting 5k or 6K, you are going to have to downsample to 4K or 2K or 1080p one way or another. Is After Effects' bicubic downsampling any better or worse than Canon's in-camera sRAW downsampling? Even if they are in the same ballpark, then the convenience of shooting only 1/2 RAW in terms of recording media and post production would be very substantial, and would get a lot of use from S35 and FF35 DSMC shooters, IMHO.

I guess another way to ask this is, once you run your full-res RAW through your demosaicing and downsampling, are you going to see similar artifacts with zone plate tests?

BTW, on a related note, is the full-sensor onboard 1080p recording option going retain some of the felxibility of RAW in terms of color temp, sharpening, etc?
 
Bicubic is pretty poor as a downsample filter goes. Box filtering is worst, and that's what sRAW is looking like to me. For the best downsampling, lanczos is pretty nice. We use something pretty special in our 4k to 2k downsample which doesn't have a fancy name, but it looks lovely.

If you do a full demosaic and downsample you will get better results.

Graeme
 
By the way, I might be eating some crow here pretty soon. I have been shooting a lot of sRAW1 (1/2 RAW) timelapse sequences under varying conditions (light, texture) with the 5D2, and some possible flaws (banding, moiré) have been popping up in some of the rendered shots. What that means is that while 80% of the sRAW1 shots might look fantastic, it appears that certain conditions might trigger artifacts, banding, and moiré patterns. If this does turn out to be the case, then as Graeme and Jim have said, it would make this function nearly useless. No one wants an unpredictable camera.

I'll update this when I get a chance to really do some more testing of sRAW vs full RAW.
 
That's pretty much why we try to be careful about aliasing and moire - you never quite know when it will show up and cause bother. Thanks for your continued input Tom.

Graeme
 
Sorry to hijack the thread. I'll give it back in a minute...:innocent:

One thing that started Tom on this discussion was the way his CPU groaned under the weight of processing the larger RAW files, one solution he saw was files of half the size. Now, he's beginning to see the flaws in Canon's technique whilst their file sizes are still massively larger than Red's but there is another possible solution to his CPU woes that I can see.

If, as Graeme has suggested, the upcoming Epic has the potential to demosaic on the fly to downsample to 1080, couldn't it also be possible for Red to build the same processing power into a PCI card to speed up the process in post? That way Red could hold onto their proprietary R3D codec and make all our lives a nicer place to be.

Who better than Red to unravel R3D files? Anyone from Red care to comment?
 
bettsy said:
If, as Graeme has suggested, the upcoming Epic has the potential to demosaic on the fly to downsample to 1080, couldn't it also be possible for Red to build the same processing power into a PCI card to speed up the process in post? That way Red could hold onto their proprietary R3D codec and make all our lives a nicer place to be.
Click to expand...

Much better to use a graphic card with some kind high-level programming (eg. Nvidia with CUDA) to do that instead of building a custom card. There's an incredible amount of processing power waiting to be harnessed for these kinds of activities available in the current generation of GPU's.
 
Anders Zakrisson said:
Much better to use a graphic card with some kind high-level programming (eg. Nvidia with CUDA) to do that instead of building a custom card. There's an incredible amount of processing power waiting to be harnessed for these kinds of activities available in the current generation of GPU's.
Click to expand...

It's a daunting task to write GPU calls which is probably why we are still waiting for any genuine help from the currently available GPU's. Whereby, building a PCI card, given that they've already written the code and have already designed the hardware to do the job in camera, should be fairly simple in comparison. And, with the cost of PCI cards being so small, who cares if we bought them now and the GPU path catches up in a year or two? Way cheaper than buying a bunch of extra CPUs to help out.
 
I just wanted to dig up this thread to say once and for all, after more testing, that sRAW is nonsense. I'm seeing all kinds of crazy artifacts in all the sRAW stuff I shot. :(

Jim, Jarred and Graeme were all spot-on correct on this.
 
Thanks for the feedback Tom. It's really appreciated.

Graeme
 
You must log in or register to reply here.
Back
Top