Welcome to our community

Be a part of something great, join today!

  • 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 :)

The answer is...

There's definitely detail there as well. Check out the little dust and scratches that get lost on the Red image.

And check out just the blue channel of the lower right square on each. It looks like there is some actual color information there that is getting lost on the Red. Or it might just be the difference in the CC between the two.

Yes, I agree that the Red AA filter + debayer mostly removes the dust and lint specks that show clearly on the Canon. The specks I'm seeing on the Canon image look like 1 or 1.5 pixel wide detail. I think Red made the deliberate choice to smooth that level of detail out to avoid aliasing, which is more scary in motion picture use than in a still frame. I'd say the Canon image has more noise and more real detail, the Red image gives you some "denoising" for free. Leaving aside blue-channel 100% crop details, viewing the image as a whole they both look very nice to me.
 
Darn, I knew it! Well I still think the Red is a gift and a wonderful wonderful thing but it's too bad that the red couldn't have little more per pixel sharpness. I could just tell on a subconscious level. I guess I'm glad that my eyes are right on the money since I can't really upgrade them.
 
...it's too bad that the red couldn't have little more per pixel sharpness.

I sense Graeme about to post here... his mantra has been pretty consistent that high per-pixel sharpness goes hand in hand with aliasing; that is to say, it is not desirable for motion-picture work, and if you want more detail then you shouldn't try to oversharpen yourself into artifacts, you just need more pixels. Which I guess is what the "Epic" is all about.
 
Darn, I knew it! Well I still think the Red is a gift and a wonderful wonderful thing but it's too bad that the red couldn't have little more per pixel sharpness. I could just tell on a subconscious level. I guess I'm glad that my eyes are right on the money since I can't really upgrade them.

I am going to post a new thread on how to almost exactly match RED with Canon, and visa-versa.

Jim
 
I hate edge enhancement. Thanks for keeping that at a minimum on the RED Jim.
 
I just came back from a 5 hour setup of a D-cinema Barco 2K. I forced the technicians to reset everything . We build the settings from the ground up. After all the geometry, focus, bare and color calibration, after me beeing a technic fetishist and huge pain in the ass... we watched Carved, the snowboard shortflic i did with Daniel Reichenbach a while ago in 2k...

We were speechless. All whites there, despite the snow and sun. 120fps. All the beauty. Cinema at its best. Daniel is a god! Build 15 downloaded via GPRS module while shooting at 9000ft.

And there it was. The truth. The new age. The ultimate freedom. The ultimate cinema system.

I ask myself, can i handle Build 16 ?
 
Canon on top.
RED on the bottom.

I personally prefer the more organic look of RED. The differences in these two cameras is by design and not accident. Even the file size has a purpose.


Jim

Ummm yeah, the organic feel of RED is way cool. I completely agree. And let this be a lesson to the rest of you...I am always right. Kidding. Or am I?
 
I myself, with what little experience I have in grading, went at the two in PS and got close to matching the RED frame with the Canon by adding just a touch of saturation, bringing it up a tad, and some unmasked sharpening. The charts got pretty close I must say.

For motion though I think many are forgetting how much more resolution is perceived while viewing motion VS a still image..thus the lower per-pixel sharpness for RED. I think most prefer an organic, smooth look anyways, that way you got lots of headroom for various looks. I think the more you know about imaging, the more you're going to appreciate what RED has done with B16, and I for one have a very limited experience in all this and am quite amazed which says something.
 
I myself, with what little experience I have in grading, went at the two in PS and got close to matching the RED frame with the Canon by adding just a touch of saturation, bringing it up a tad, and some unmasked sharpening.

You can also get the Canon to look a little closer to the Red by just adding a 1 pixel Median blur. :whistling:
 
Yeah, build 16 will rock my world.
Build 15 is great already but 16 will kick RED in to overdrive.
I cant think about the Epic.
Keep on going guys. It's so much fun to work with this baby.

Cheers
Tim
 
Yes, JBeale is right... You've got to be careful about two things for motion images that you don't have to worry as much about for stills - aliasing and temporal motion judder. We don't use the weakest optical low pass filter. We don't want that level of aliasing on moving images.

But what about judder - well, in a way, the perception of judder is related to edge sharpness. I think that's one of the reasons why HD judders more than film - the other being the perception of DOF where shallow DOF helps with judder, which is harder to achieve on 2/3". Well, edge sharpness goes hand in hand with aliasing, and with how the demosaic is done. It's one thing to design a demosaic that works great for stills, but I think you need extra care, especially with something adaptive, for motion. It needs to be completely tolerant to noise, and it needs to work well with the REDCODE compression. You can't just put an off-the-shelf demosaic on an R3D and expect the same level of results. When I downloaded Jim's two images, yes I can tell which is which. However, I did the "standard" test I do, which is crank up the saturation to see which image "breaks" first. Well, what I saw was even with two applications of hyper saturation, the RED didn't break up as much, to my eyes, as the Canon. Next, I noticed that at the edges of the image the blue channel in the Canon had a ghostly blur, absent from the RED, and some of the other colours seemed to glow more too.

Graeme
 
I read the histogram in Photoshop and Canon's image had more data in it. For printing matters, even though the Chips are different, one image would probably be 72ppi and the other 300ppi (RAW) so the sharpness issue would be quite different on a 100% crop.
The trick about the DoF was genius, but when you check the channels you can see that red's has a little chromatic aberration on the edges near the BASICS acrylic paint cup. Canon lenses correct this almost perfectly. Unfortunately I still havent got my red to try them side by side. Nice exercise.

Still patiently waiting...
 
Canon on top.
RED on the bottom.

Jim, that is amazing.

I'd be interested to see the 5K shots versus DSLRs when the time comes.

I'd dump my D3, 5D, SD14 and HVX200 in favor of it!

Very interesting.

95% of my image delivery is for the web. I'm usually reducing images by 80% for delivery anyway.

Wow... I'm just blown away by that image comparison.
 
quantitative color analysis

quantitative color analysis

I had a quick look at the colors in these comparison using Imatest 2.7. Since the shots include a standard color chart, you can do some numerical analysis on them. Imatest reports the Canon image to be closer to the standard color values (assuming both JPEGs are in sRGB space). I assume the matching could be improved by appropriate color matrix transforms, if you wanted to.

If this chart doesn't make sense to you, don't worry- it's more of a geeky tech analysis, I doubt many working cinematographers think about color this way. But as a quick summary, if you had perfect camera generating a perfect sRGB JPEG, you'd see the circles and squares exactly overlap everywhere. The longer the line between circle and square, the greater the color error. The center of the chart is zero saturation (black, white and greys). You see the Red values tending towards the center, so they are less saturated than the ideal (mostly) but even the Canon is not quite as saturated as the ideal chart values. One of which (yellow square #50) is actually outside the sRGB gamut, so it cannot be accurately represented without moving to a larger color gamut. The Red's version of some of the neutral grey tones also drifts a bit towards blue-magenta.

(sorry I have to include the chart as a JPG. The PNG version is much smaller, but the PNG filetype is not allowed for upload here.)
 
(assuming both JPEGs are in sRGB space).[/SIZE]

Is it possible to do the same analysis on a RAW untouched frame from RED? Or would that not work- sorry the tech stuff sometimes eludes me. (to me- RAW = uncooked; JPEG = just poached egg; TIFF = something I do when I'm angry; PNG = a sound a submarine makes :clown2: :clown2: :clown2: )
 
(sorry I have to include the chart as a JPG. The PNG version is much smaller, but the PNG filetype is not allowed for upload here.)

hey.. at least its free :)
 
Not that it will change the results, I grade in Adobe 1998. While there were a few things to skew this test (like changing colors in a haphazard manner to make them look more similar... and not doing a very good job of it) these kinds of reports are very healthy. We are absolutely certain that we can dial color in better. Actually, we assume that whatever we are doing can be done better. That is probably why we are advancing things along so quickly. Thank for the work.

Jim
 
I have to be honest here: I didn't see this type of image quality coming for the next 2 years, not from anyone. That comparison is worth more than all the marketing in the world.

The RED has really put a spin on things here in SA as well. Some love it, others hate it. I showed it to a guy who used to work for Sony and has over 30 years tech experience behind him and has not been very warm to the idea of RED, and when we switched it on he said, "Oh my..." and smiled for the rest of the day.
Well done!
 
Back
Top