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 :)

Scarlet Resolution

Michael Dearworth

Well-known member
Joined
Dec 25, 2010
Messages
185
Reaction score
0
Points
0
I know RED has stated in the past that other cameras that are labeled as "1080p" are really somewhere around 720. Does this rule apply for RED cameras as well? Is 3K on the Scarlet really somewhere around 1080/2K? I'm planning on buying this camera, so I was just curious. Thanks
 
Three factors

Three factors

I know RED has stated in the past that other cameras that are labeled as "1080p" are really somewhere around 720. Does this rule apply for RED cameras as well? Is 3K on the Scarlet really somewhere around 1080/2K? I'm planning on buying this camera, so I was just curious. Thanks

The "Resolution" from any digital camera using a Bayer filter needs to take into account three factors:

1) The lens resolution which in most cameras is less than the sensor pixel resolution at 3K, but more so in smaller sensors because of diffraction, optical aberrations, and the OLPF+IR/UV cut being thicker in ratio to the sensor active area height.

2) The general limits of grid shaped pixels in a digital image, all digital images will alias if you try to get 100% edge contrast to the single pixel level.

3) The quality of the de-Bayer, the over-sample ratio, and how high the OLPF compensation sharpen is set. You can vary the apparent "resolution" over a wide range by adjusting the OLPF compensation sharpen value in the de-Bayer processing, in fact you can get more than 100% MTF that way.

Without ANY OLPF compensation on a small sensor camera with average lenses the visable "resolution" would look quite low.

There is no fixed value other than the MTF roll off should be such that you are less than 50% MTF at about 0.707 the image height, 1080*0.707 = 763, but that is true for any camera.

The 2880x1620 oversample for 1920x1080 is like ARRI uses in ALEXA, so it gives an advantage in reducting some Bayer artifacts and the blur from the OLPF.

For use as a Cinema Camera, the dynamic range and how well the camera renders color and image tones, and the lack of compression artifacts can be more useful than a test chart resolution value that can be tweeked up or down with the OLPF comp value. You need to look at the results. If you are going to 35mm filmout, Blu-Ray, or DVD what you can get away with will vary since any re-size of the image or transfer to film could wipe out small resolution differences.
 
I think they said it will be close to 2.4k, but I could be wrong I'm sure someone will pipe in and correct me if I'm wrong though I doubt many if any outside of red know the actual number.
 
Measured resolution on Scarlet should be significantly better than 2k. If we assume for arguments sake the measured 3.2k from RED One's 4k capture, that would be ~2.4k for instance.

And that is why you shouldn't start with 1080p if that's what you wish to achieve, and if you have more than 1080p on your sensor and that's what you're aiming for you'd better deal with all those pixels properly and in full.

Graeme
 
Thanks so much for the responses, guys. That answered my question.
 
I think they said it will be close to 2.4k, but I could be wrong I'm sure someone will pipe in and correct me if I'm wrong though I doubt many if any outside of red know the actual number.

Measurements have seemed to be consistently 80% of original resolution.

EPIC 5K = 4K debayered
RED One 4K = 3.2K debayered
Scarlet 3K = 2.4 debayered
 
I know RED has stated in the past that other cameras that are labeled as "1080p" are really somewhere around 720. Does this rule apply for RED cameras as well? Is 3K on the Scarlet really somewhere around 1080/2K? I'm planning on buying this camera, so I was just curious. Thanks

Also, another reason why these cameras you are referring to are labeled 1080p is because after they resolve almost 720p from a large sensor, they are then stretched right back up and wrapped in a codec that is set to 1920x1080.

It's like "If you aren't screwed enough to lose that resolution, lets encode it to the compression YouTube can handle. If that's not enough, let's artificially scale it all the way back up to 1080p.

And then the consumer relatively says "We can't get enough of the junk...if you can make it cheaper, give us more junk!"
 
The "Resolution" from any digital camera using a Bayer filter needs to take into account three factors:

Of course, it's worth noting that most professional 2/3" cameras are using three sensors and not a Bayer-filtered single sensor.
 
Similar issues occur on 3 chip systems too - generally gaining their resolution through allowing too much aliasing.

Graeme
 
Advertising has done a great job of telling people that "Full HD 1080p" is better than everything else... even if the 1080p their camera is recording is actually storing less resolution than standard-def.

Case in point - VDSLRs that record 720p automatically being considered inferior to those that record 1080p, even by high-end review sites, who of course never bother testing what they actually resolve,
 
So, when you have a DSLR that is recording video, shouldn't you be able to get full 1080p resolution? Since the camera is downsizing from a much higher resolution RAW image coming off the sensor?

I know there are all kinds of problems with a lot of the DSLRs related to moire, aliasing, line skipping, and/or binning. Why can't a DSLR that takes a 16 MP RAW image and scales it down to 1080p get full 1080p resolution? For example, it appears that people are generally agreeing that the Panasonic GH2 is the best video DSLR available right now (well, if you can find one). I can't find anything where anyone has actually tested the resolution, though. Assuming it's lower than 1080p, is it the codec and bit rate that is limiting it's resolution? (I probably just answered my own question there, huh?) And why are the other ones resolving only 720p if they're coming off a much bigger sensor?
 
The sensor size is not what gives it the resolution...I think the Epic, at 5K does like 18MP or something and the Scarlet at 3K is supposed to be like 5MP...
 
With any camera system there's a balance of qualities - high resolution, low aliasing / moire, and low ringing / sharpening artifacts. If you want maximize any one quality, you may end up sacrificing on the others.

But to even begin to get into that situation you need to be reading all the pixels individually off the sensor. If you skip rows or bin pixels you will already be heading towards aliasing and moire and that is un-recoverable by future processing.

Now if you've read all the pixels, which is doubtful in these vDSLRs, you still have to generate a 1080p image. If that is not done properly (as in expensively in terms of processing computations with the inevitable heat and power requirements, not to mention cost) you should be able to balance the resolution / moire / ringing qualities to produce a very nice 1080p image!

A camera can fail at any of these stages - it can take shortcuts all along the route in the hope that the end user won't see them, or won't notice the issues.

As mentioned, the codec can indeed impair resolution above and beyond what has been lost by the earlier shortcuts in the pipe, but a fully detailed 1080p image produced by proper sensor reading and downsampling would make life very hard for a codec by presenting it with a lot of detail. There's a lot of factors and they're all intertwined.

Graeme
 
So, when you have a DSLR that is recording video, shouldn't you be able to get full 1080p resolution? Since the camera is downsizing from a much higher resolution RAW image coming off the sensor?

If they were actually able to read out an 18MP image and scale it down 24 times per second they would be able to resolve much closer to 1080p and would not have as many aliasing problems. The problem is that they don't have the processing power to do that and that's why they resort to tricks like line skipping and pixel binning.

Also note that even with downsampling you are never going to be able to completely resolve the format resolution. Downsampling itself creates aliasing (recording an image in pixels is basically downsampling from infinite resolution). To combat this you have to blur the image before sampling to remove frequencies that are going to alias, and so you lose some resolution just like you do with an OLPF.
 
In other words Scarlet hopefuls... there is more to Scarlet than excellent resolution/$

but also Graeme's Magic (REDCODE:super sexy wavelet compression algorithms) and REDCO Color Science
to create what Im willing to bet is going to be some of the tastiest footage and features for your $ - hands down.

Time will tell
 
Thanks for this thread. This is some of the most calmly explained and detailed reasoning as to why the Scarlet will kick ass over the competition, that I've read here on Reduser. I wish all threads could be like this over here, rather than just bashing the DSLR's with no explanation.

Thanks for the info everyone.
 
In other words Scarlet hopefuls... there is more to Scarlet than excellent resolution/$

but also Graeme's Magic (REDCODE:super sexy wavelet compression algorithms) and REDCO Color Science
to create what Im willing to bet is going to be some of the tastiest footage and features for your $ - hands down.

Time will tell

I am so looking forward to a camera that will more likely be limited by the lens I put on it or the output format and codec the footage ultimately gets crunched too for release than the image capture capabilities of the camera itself.
 
Hi Graeme,

Thanks for stopping by to answer a question or two. Your input is greatly appreciated as always.

I was wondering if there are any plans for a 3.5K mode on the Scarlet akin to the 4.5K mode on the R1 (i.e using the look around area)?
 
Thanks for the detailed response Graeme. I hadn't realized that the line skipping and pixel binning meant that they were actually not reading all of the sensor (probably at least, I have a GH2, and I'm not sure anyone has figured out exactly what they're doing). That makes it all make much more sense!

I just love that REDCODE is RAW. None of the DSLRs even come close to what Scarlet is going to be able to do, with RAW, 3K, and HDRx™. I'm most excited to see the quality of the stills that can come out of Scarlet. I've been a stills guy, and I'll keep my DSLRs around for for stills, but rarely do I have a need for a full 10 or 16 mp still image for the work I do, and I can't wait to see the quality of the stills that come out of Scarlet Fixed. Especially at low light, where the effective resolution of the GH2 is far far lower than the 16 mp of the sensor.
 
Back
Top