Comparison of Native 4k, Native 1080p, and Bicubic Downsampled images

Musicvid wrote on 1/28/2018, 6:03 PM

In that order.

A variety of original unadulterated camera tests are solicited.

Critiques without supporting visual comparisons are afoul.

Comments

3POINT wrote on 1/28/2018, 9:57 PM

Nice example, view your picture from a normal viewing distance and tell me which one looks almost as crispy as 4k?

3POINT, Theo Houben, Vegasuser since version 5

Recware: DJI Osmo Pocket, GoproHero7Black, PanasonicFZ300/HCX909, Sony PXW-X160.

Software:VegasPro17/VegasMovieStudio16/PowerDirector18

Hardware:i7 4770, 16GB, GTX650ti, 1920x1200 display

Playware:Samsung Qled QE65Q6FN, Philips 55PFL7108

Musicvid wrote on 1/28/2018, 10:32 PM

Define "normal viewing distance," on what size screen, and then you tell me.

I'm not the commentator at this ball game.

OldSmoke wrote on 1/29/2018, 12:16 AM

Which camera did your 4K and 1080 image come from? Using a graphic and down sampling it is not the same as using an actual image of a scene in two different resolutions.

Proud owner of Sony Vegas Pro 7, 8, 9, 10, 11, 12 & 13 and now Magix VP15&16.

System Spec.:
Motherboard: ASUS X299 Prime-A

Ram: G.Skill 4x8GB DDR4 2666 XMP

CPU: i7-9800x @ 4.6GHz (custom water cooling system)
GPU: 1x AMD Vega Pro Frontier Edition (water cooled)
Hard drives: System Samsung 970Pro NVME, AV-Projects 1TB (4x Intel P7600 512GB VROC), 4x 2.5" Hotswap bays, 1x 3.5" Hotswap Bay, 1x LG BluRay Burner

PSU: Corsair 1200W
Monitor: 2x Dell Ultrasharp U2713HM (2560x1440)

Musicvid wrote on 1/29/2018, 2:51 AM

I was kind of expecting you would show us that, OldSmoke.

Didn't your "actual image(s) of a scene in two different resolutions" upload correctly?

3POINT wrote on 1/29/2018, 5:42 AM

Musicvid, do you have a 4k camera? Just do the test by yourself. Tip: use a scene with a tree with lots of leaves waving in the wind.

Last changed by 3POINT on 1/29/2018, 5:43 AM, changed a total of 1 times.

3POINT, Theo Houben, Vegasuser since version 5

Recware: DJI Osmo Pocket, GoproHero7Black, PanasonicFZ300/HCX909, Sony PXW-X160.

Software:VegasPro17/VegasMovieStudio16/PowerDirector18

Hardware:i7 4770, 16GB, GTX650ti, 1920x1200 display

Playware:Samsung Qled QE65Q6FN, Philips 55PFL7108

Brandigan wrote on 1/29/2018, 5:48 AM

As I said in the thread you forked this from: it doesn't work like the B/W test images above. 3Points's suggestion is a good one.

Nick Hope wrote on 1/29/2018, 6:19 AM

As I said in the thread you forked this from...

https://www.vegascreativesoftware.info/us/forum/4k-monitor-is-downscaled-hd-from-4k-sharper--110129/

Kinvermark wrote on 1/29/2018, 7:21 AM

Camera test images would be good. But in fairness, it is not just up to Musicvid to prove his point, others must also prove theirs - otherwise it is just an unsubstantiated claim. I am on the fence with "it depends" as my answer. I "think" that the GH4 UHD down-sampled was better than HD native, but I am not so sure about the GH5. I am really busy right now but will try to get a samples up when time allows.

Can we agree on a simple workflow (using Vegas of course)?

1) HD 30p on HD timeline rendered to HD 30p - 1080 Still image grab.

2) UHD 30p on UHD timeline rendered to HD 30p - 1080 still image grab.

Subject material suggestions? - not outside, it's raining:) ... and dark.

Brandigan wrote on 1/29/2018, 7:35 AM

Camera test images would be good. But in fairness, it is not just up to Musicvid to prove his point, others must also prove theirs - otherwise it is just an unsubstantiated claim.

He's not making a point, he's closing his eyes and saying "I can't see anything", then starting from a different premise and asked for it to be proved wrong. No one has to prove anything. Google and you'll find all the proof you want, but that's apparently still not enough. Those of us who know it works are already reaping the benefits, so we're fine.

Try what I suggested in the other thread. It's not practical to work from two images taken even a fraction apart. Someone will always complain the camera did blah blah, or the light changed, or the trees moved in the wind etc. etc. Work from the same, single, known large image as a stand in for reality and manipulate it in the two ways the camera (and then subsequently you in Vegas) would and examine the results. The camera used, or the image, is immaterial. It's the process that's relevant. No sleight of hand involved. It just works.

Kinvermark wrote on 1/29/2018, 7:44 AM

Nonsense. Just show us YOUR proof. The images will be close enough to make a reasonable judgement. I didn't say I don't believe you; I said I have an open mind. I don't want a "stand in" for reality. I want actual camera video stream capture as a starting point. I will post images when it's not dark here.

PS - Please stop attacking Musicvid. Not relevant . You're just undermining yourself.

Former users wrote on 1/29/2018, 7:45 AM

Two 2K frame grabs for comparison from GH5 shot at max data rate for original 2k and 4k

The top image is from a rendered out 2k .mp4 at 28mb/s, from 2K 200mb/s original all-1 GH5

The bottom image is from a  rendered out 2k .mp4 at 28mb/s, from 4K 400mb/s original all-1 GH5

So for clarity i'm not including the original 2k or original 4k file frame grabs, after all these 2 above are what you might be delivering.

If you want to do your own testing with the originals i’ve put them here ...

https://www.dropbox.com/sh/j6p4adbm983p435/AABq-0tGLVRzqRDHnAufINjIa?dl=0

Kinvermark wrote on 1/29/2018, 8:05 AM

Thanks JN_. Except for the shadow passing over in one shot they look identical. Even zoomed in on an UHD monitor. Just curious - which lens?

Brandigan wrote on 1/29/2018, 8:17 AM

Nonsense. Just show us YOUR proof. The images will be close enough to make a reasonable judgement. I didn't say I don't believe you; I said I have an open mind. I don't want a "stand in" for reality. I want actual camera video stream capture as a starting point. I will post images when it's not dark here.

PS - Please stop attacking Musicvid. Not relevant . You're just undermining yourself.


If you think it's nonsense, then you probably wouldn't know what you'd be looking for anyway, so maybe you're better off not bothering.

Only your own proof using your own methods will be enough. I have the results already, thanks. We've explained how it works, You want it? Some work on your part is involved.

A huge image as a stand in for reality is the only way you can veritably reproduce a difference in the result from the two methods. It's those which are the important thing, not the source. It works with anything.

I'm not attacking Musicvid, just pointing out that it's actually easier to find the proof and try it yourself than create examples where it doesn't work, because: Maths.
 

Former users wrote on 1/29/2018, 8:21 AM

Yes Kinvermark, the shadow😊. My own preference is for the downsampled 4K, you have to look close but the clarity improvement is there imho.

I used the “kit” lens purchased with it, a saving of near enough €300 that I couldn’t resist. The 12-60mm Leica branded Pana. Maybe check out the .jpg's that i’ve uploaded to dropbox, its easier to see, but not that much.

Brandigan wrote on 1/29/2018, 8:25 AM

Thanks JN_. Except for the shadow passing over in one shot they look identical. Even zoomed in on an UHD monitor. Just curious - which lens?

They're not even close to identical, because they're taken from two different angles (the plants move,) and one is using a flash. 😆

Even so, if you zoom into the chart you can see some differences, although as the camera had moved, can't tell what to attribute those to.

Needs to be a single image.

Former users wrote on 1/29/2018, 9:22 AM

There was no flash used, that was the Sun doing its thing. The lens was set to aperture priority f4.0 and manual focus. The test is good enough to show the slight difference in clarity, which imho gives the nod to the downsampled jpg. By all means do your own test, no ones holding you back.

Brandigan wrote on 1/29/2018, 9:51 AM

There was no flash used, that was the Sun doing its thing. The lens was set to aperture priority f4.0 and manual focus. The test is good enough to show the slight difference in clarity, which imho gives the nod to the downsampled jpg. By all means do your own test, no ones holding you back.

Your test proves exactly nothing that this subject is about because the images are so different it is pointless to compare them.

A better comparison would be to at least try and take the images in the same 1/4 second and don't move the camera. At all. Not even a little bit. Can you control it through an App on a smartphone?

I've been doing my own "tests" for years, my clients love the results, thanks. 😉
 

D7K wrote on 1/29/2018, 9:57 AM

From what I've read it really depends on a whole list of factors, and number one is does the camera actually capture 1080 lines of resolution on the vertical when shooting 1080P. You of course do get 1080 lines when you down rez 4K. I have two type of cameras I use for video. The Gx8 and Gx85 - the Gx85 does less sensor cropping than the Gx8, and I think for video only the Gx85 might produce better 4K to start with. I happen to like the Down Rez'd 4K better but that is because I know neither camera does 1080 lines of vertical resolution when they shoot 1080P. Heck at 68 on either my 1080P or 4K TV's I'm just glad I can see the video, and on a phone or tablet, honestly who cares? Theater relase - then big deal probably. There are so many different systems for viewing, its all a crap shoot now-a-days IMHO. It is never a good thing to turn personal in a forum IMHO.

Musicvid wrote on 1/29/2018, 10:12 AM

My goodness!

1. I made no claims. Only provided a baseline example, which is being shredded by emotion, and not by counterexample.

The original post contained only a request for media and a caveat. Only one person has come forth with camera images of ANY kind, and jn_ has scored the first base hit in a game regretably overrun by foul balls and errors.

2. Yes, this is a "show me yours" thread, because my real world tests would be on a GoPro, which no one thinks passes muster!

3. No cameras were harmed in the making of my demonstration. The frst two (native 4k and 1080p) are obviously generated from the same pattern, the third was bicubic downsampled from the 4k original, and should provide some clue as to whether the down-dithered image is "sharper" than 1080p natively in a sterile setting. Other forms of downsampling are available for testing inside Vegas; its unfortunate Lanczos isnt one of them.

4. A couple of individuals are engaging in unnecessarily harsh critique (again, without ANY first-person evidence whatsoever.) Thats unprofessional, inappropriate, and i wont tolerate it. Go play on your own damned sandlot, or risk a complaint. Its that simple.

5. As recently as a couple of years ago, forum protocol more or less required a contribution of something more than hot air to be taken seriously. I realize we are dealing with a younger, less critical, and perhaps more impulsive fan base overall than in the past. For that reason, the task of preserving a scientific method, pointing out magical thinking and internet hype for what it is, and reminding others that we all began learning from the same place has fallen to a relatively few stubborn oldtimers, and some remarkably gifted 2nd gen types who didnt have to cut their teeth on an analog model.

My real hope for this thread is that people with the time and equipment will start posting unambiguous real world tests of their capabilities at varying resolutions, and allow comparisons with post-acquisition downsampling, which is being touted as some kind of magic bullet, regardless of inherent losses and blurring that all forms of dithering produce.

But no, a few have instead shat all over an otherwise sentient investigation, proving once again that entropy is the most powerful force on this planet. And we all know where that's gotten us. Not pretty.

Carry on...

Brandigan wrote on 1/29/2018, 10:16 AM

Well, that's an interesting argument. So, let's consider it.

The manufacturer says the sensor can capture 4K, but maybe they're lying and it only does 3.5k because (no idea. Fun? Marketing BS?), which while not full 4K, is pretty good and will definitely give you 1080 lines when you downsample it at least.

But that same camera does not use the 3.5k sensor to produce 1080p lines for some reason? Why? What do you think it does? Just misses out every alternate line and gives you 1.75k, or 1000p or something? How do you know this?

But what about those cameras that do produce full HD 1080p in the first place?
Glad you asked. It doesn't matter, the downsampled 4K will still look better, which is where we came in. 😆

Nick Hope wrote on 1/29/2018, 10:25 AM

Folks, be cool. Before this heats up any more, you are all reminded of community rule #1:

1. This is a community – not a competition

Respect the opinions and views of other members of the community.Insults, coarse language, and harassment of any kind toward other community members is forbidden. Avoid provoking other members and keep your comments factual and objective. Angry posts written to blow of steam will be deleted. Members of the community who write such posts will be warned. Repeat violations will result in removal from the community.

Brandigan wrote on 1/29/2018, 10:27 AM
 

2. Yes, this is a "show me yours" thread, because my real world tests would be on a GoPro, which no one thinks passes muster!

Actually, the camera doesn't matter, so it even works on a GoPro, and I said as such previously - having been one of the first cameras to do 4k video at a bargain price, even if it was only at 15fps. So no excuses for not trying it yourself now.

The title of this thread is confusing, as Native 4K obviously looks better on a 4K screen, so will look nothing like either Native or Downsampled 1080p. The comparison is only between Native 1080p and Downsampled 4k>1080p.

Same resolution, more colour information. That's all it is.

No dithering -which only you mentioned - is involved whatsoever.

Whatever your age: Google is your friend. Treat it well and it will reward you.

Former users wrote on 1/29/2018, 10:33 AM

“A better comparison would be to at least try and take the images in the same 1/4 second”

I can understand your frustration with all of this, unfortunately, my time machine is in for repair.

Since I shot each of the 2 clips about 10 seconds long, on the same camera system, and picked an approximate mid point for the frame grab, the same point for every clip, I really do need to complete the repair first to get that 1/4 second gap.

Alternatively, if you can accept say an approximate 15 second gap I can redo the test tomorrow, its 4:30 as I write this and will soon be dark. As to the Sun, and stopping its movement, even I, with a broken time machine cannot help.

.

“Your test proves exactly nothing that this subject is about because the images are so different it is pointless to compare them.”

Actually if you look at the 2 images, although the artificial flowers have moved over, everything else in the image, position wise, is identical. The light changed but I really dont think it matters in evaluating the 2 images for clarity.

Unless someyone else requests a new upload i’ll let it stand.

Musicvid wrote on 1/29/2018, 12:00 PM

Brandigan, im willing to engage in a discussion with you, but only if you provide first-person evidence to support your claims. Otherwise, for the purpose of this discussion, its all just background noise. I have provided only a demonstration and have made no claims whatsoever, except for the one that bicubic dithering in post is a form of blurring, not sharpening. Dithering is an integral function of downsampling. If you want to argue native vs. simple scaling, start your own thread because im just not.interested. That said, i am certain that some cameras will fare worse than others at internal 1080p, even to the the point of the post-downsampled image "looking better," your words.

But thats not the question, is it? The question is and remains whether the downsampled image is SHARPER. Sharpness can be measured, and yet your dramatic insistence on equivocating the qualitative "better" suggests that you haven't gotten the point of my discussion, at all.

Same resolution, more colour information. That's all it is.

Perfect example. Native 1080p and downsampled 1080p are both presumed to be YUV 8 bit 4:2:0, so the "amount of color information" (number of colors) is the same, save for losses from downsampling itself, and the acquired 4k is already starting with less available color information than the camera sensors had on-scene. The color information that is decimated in the conversion from REC2020 is nonrecoverable, no physical exceptions. You can't unbake that cake. So if you still believe that premise without providing any supporting tests of your own, i implore you to share it in that first thread, where you seem to have acheived a measure of hero status, and where some of your peers already seem to be claiming to turn lead into gold, and perhaps mount a crusade. I feel blessed to have escaped in time.

Seriously, i wouldnt be spending this kind of time if i didnt recognize some potential for you to become a thinking and informed contributer. I'll tell you a little secret -- the times Ive been caught red-handed dishing misinformation to newcomers are always the times i haven't bothered to pretest my theories, down to the last detail. And each time i am reminded of my own responsibilities of credibility on a platform that is seen by many.

As a one-year contributer, i welcome you to the extended Sonic Foundry / Sony / Magix forum for Vegas professionals, to which ive contributed (?) since '02. You've shown some original thinking, which is always recognized.

The comparison is only between Native 1080p and Downsampled 4k>1080p.

You have my complete, unmediated agreement.

Now, about those tests..................................................?