Why is Neat video the best video denoiser?

[Truthler]

I think he's talking about some old DSLR limit in which you can record only for 30 minutes.


Also @truthler, try changing the lens on your camcorder, I dare you

Why shld I cahnge? You can not attach true professional lenses to your DSLR....

Try to connet that to your still camera: https://www.3dbroadcastsales.com/canon-cj18ex7-6b-kase-4k-uhdgc-portable-eng-broadcast-lenses

Its cover comes off from the weigth. Still cameras are NOT ideal for outdoor filing, due to their shit form factor. Let's don't forget the small shit individial pixel sizes of its sensor and the heavy denoising processes...hahaha

Simply they were not designed for outdoor.

That is wrong. Quite the opposite in fact. Yes, you can attach pro lenses to dslr. And there is a whole line of converter/adapters built and sold to allow for it and all its brand & size variability.

https://www.google.com/amp/s/www.bhphotovideo.com/explora/amp/photography/tips-and-sol...-lens-adapters

Weight can be an issue. But you need to think outside of the box and realize that the dslr does not need to be the object that is solely doing the supporting. There are support sleds for those ridiculously large and heavy lenses that hold the adapter rings, and those do the job of holding up BOTH the lens and the camera.

But how about you get back on topic for this thread and provide those capture steps, intermediate steps, AND SETTINGS, in DETAIL, that we have requested from you multiple times?

Scott

But these guys want to solve the problem with cheap budget. And Let's don1t forget, that you always must buy 2 cameras/camcorders to make different angle shots. Two device is the minimum. They had to sped some thousand dollars to compete with the ZEISS lens system of the AX700. If you want to buy lenses for still cameras, you must spend 2X 3X more money to achieve the top consumer model camcorsers lenses.

Moreover AX 700 is not broadcast camcorder, only because it is not provided an upper handle which contain the XLR audio box of its expensive brothers (like the Sony PXW-Z90 ).


But you can buy additional audio system for AX700 with XLR audio box.


See: https://youtu.be/gPziwbitbu4

And with Cam Link 4K, you can really shot lossless quality.

About weight: Until these guys achieve the tolerable quality lenses , the holder, the external Audio etc... the weigth of their stuff can reach 2-3 Kg )))

[Truthler]

You can't build a good comparable Still camera set under 3000-3500$

Proof, watch that video: https://youtu.be/Jgk8FOF8JbM

Even the audio elecronics is a scrap in still cameras, so it is not enough to buy a Good microphone, but you need external audio recorder. A comparable quality zoom lens system const 1500 - 2000 $.

Still cameras had not enough dedicated buttons for various functions, so their usage is more difficult.

And again: Let's don't forget, that you always must buy minimum TWO machines if you deal with video.

[Truthler]

https://www.techradar.com/reviews/samsung-galaxy-s21-ultra-review

Samsung S21 phone shot 8K video, which is better than most of still cameras which usually shot only in 4K....

[rrats]

But I did not speak about astro-photgraphy, but about average quality (often no name without old brand) consumer still cameras what these guys want to buy. Their consumer grade still cameras were unable to compete even with an Iphone 12 in video quality ))

my favourite no name without consumer still camera brand is canon

[rrats]

Truthler, are you really telling me Samsung galaxy s21 makes better video that most DSLR/mirrorless cameras?


It's 8k videos suck a-- in my opinion

[Truthler]

Truthler, are you really telling me Samsung galaxy s21 makes better video that most DSLR/mirrorless cameras?


It's 8k videos suck a-- in my opinion

What is your problem with 8K videos? Higher resolution means better quality and less halo effect. Halo effect is always there on every time for a some degree when you make photo or video which contains high contrast areas. (Just enough to make a magification to see it) However it is less noticeable, but if you give sharpness it became more obvious. In higher resolution 8K and 16K it is less relevant effect, because there are so many pixels, where some pixel wide halo will be not so noticeable...

[rrats]

and less halo effect

you were the one adding sharpening and providing us sharpened camcorer footage

Halo effect is always there

its not. you will get it tho by using come crap like 4:2:0



The thing I was trying to say is that samsung s21 records awful looking 8k videos. My mirrorless's 4k would easily beat it

[Truthler]

and less halo effect

you were the one adding sharpening and providing us sharpened camcorer footage

Halo effect is always there

its not. you will get it tho by using come crap like 4:2:0



The thing I was trying to say is that samsung s21 records awful looking 8k videos. My mirrorless's 4k would easily beat it

Your cheap mirrorless can not make good quality video. Their video often did not reach an Iphone 13 level. You must buy around 6-8000 $ still camera to beat an Iphone 13 or Samsung S21 phone.

How do you feel when it comes to your mind that even good quality mobile phones shot better videos than consumer grade DSLRs and mirrorless cameras?

[rrats]

Do you seriously believe in that crap?

[Truthler]

Do you seriously believe in that crap?

Yes. I am convienced that consumer grade DSLR and mirrorless cameras can not compete with high quality Mobiles. You must have a deeper pocket (professional level still cameras) to beat mobiles!

Just an addition, Some of hackers already working on firmware modification, and the Ax700 get 4:2:2 10bit MXF recording format like their more expensive brothers, because their HW and electronics are exactly the same.

[rrats]

I am convienced that consumer grade DSLR and mirrorless cameras can not compete with high quality Mobiles

You must be pretty delusional then. Phones have small senors and small lenses. they have no real zoom that you can freely ajust

explain to me how this looks good in any way: https://youtu.be/2tbPi2_aBms

[Truthler]

I am convienced that consumer grade DSLR and mirrorless cameras can not compete with high quality Mobiles

You must be pretty delusional then. Phones have small senors and small lenses. they have no real zoom that you can freely ajust

explain to me how this looks good in any way: https://youtu.be/2tbPi2_aBms

Phones have way more intelligent softwares than still camcorders due to their stronger more advanced computig capability.

Do you know that Samsung and Iphone already uses (yes primitive due to the HW capabilities) artificial intelligence to create enhanced quality videos and photos?

Even the macro is better with Iphone max pro

See that for proof:

https://youtu.be/jHEfdyKF4Ws

See what professional filmakers says about Iphone 13 pro max.

https://youtu.be/c3CZX-lnAIc?t=170

Yes, with addition it can record in prores too.

With expensive additional softwares it can do "on-the fly" pre- color grading and more color profiles than DSLR-s or mirrorless cameras will ever have.

[rrats]

ai != physics

See what professional filmakers says about Iphone 13.

it's obvious hes telling good things about that lighting-powered crap when its on apple official youtube channel

[johns0]

Thread moved to off topics since its no longer concerned with encoding.

[Truthler]

ai != physics

See what professional filmakers says about Iphone 13.

it's obvious hes telling good things about that lighting-powered crap when its on apple official youtube channel

How interesting, you forget to react this video because of obvious reasons.

https://youtu.be/jHEfdyKF4Ws

[poisondeathray]

It would be more interesting to post a response video with SM-Degrain (I guarantee that it will be worse), and you can write chriticism after that.

SMDdegrain alone? What happened to team open source? Are you limiting it now ?

Worse than what? I guarantee it won't look as bad as the mess you posted if you allow team open source to compete.

And how can you "guarantee" anything when you haven't used any of the filters properly (including Neat Video) ?

The way people use Neat and SMDegrain , just basically default settings, - in general I think Neat will produce better results and faster. But it's not the BEST results. Neat might be good for general user, but it does not produce the best results for power user. Neat is the editors choice because it produces good results, fast. Editor doesn' t have time to invest for the best results.



If you use SMDegrain + pre stabilize, pre/post filter, it will produce similar or better results in general to Neat Video. But much slower - in the order of 20-30x slower (!) for the denoising step itself. SMD will produce noticably better results on sections like the building, newspaper. Why? Because the TR for the avs+ version is not limited to 5 like neat video. For "regular" scenes with moving people/objects, that you prestabilize, the difference will not be as large (because you use a lower TR to reduce ghosting), but open source will still produce similar or better results with less ghosting. I mentioned it earlier, the temporal denoising is still a slight relative weakness for Neat V5. At large TR values, SMD approaches the quality of stacking methods, but SMD does not have the contamination issues with large peripheral borders, or as large issues with FG ghosting compared to mean/median stacking. Aligned areas or frames will still be the best with stacking methods using mean/median

Understand that I'm using "lock frame" stabilization - because it produces the best results for any denoiser using temporal denoising, including Neat Video. Each "n" +/-x frame is stabilized on that "poster" frame, the single output frame kept, the rest discarded. Repeat next frame. Ie. each current frame requires a set of stabilized frames in reference to that "poster" frame. You can automate some of the tasks with scripts and keyframes. This is why it's complicated and takes a long time. This way, you don't get peripheral border contamination for small to medium borders, even with Neat. But the question was about "best". This is the best method, even on scenes with moving objects (you can use slightly lower TR to reduce ghosting). And if you use everything properly and include complicated filter chains, it's difficult to compete with a full array of options of open source. This is the reason I don't post full video - it takes too long to get the absolute best - you'd only do it "for real" .

I didn't post the newspaper using smdegrain with large TR, but the results are clearly mean stack >> smdegrain > neat in terms of signal to noise ratio. (And there are no tricks I know with Neat Video that can improve it more)

AVISource("445-506_stab_ut_10bit422.avi")
SMDegrain (tr=20, thSAD=1200, thSADC = 800, refinemotion=true, mode=0, contrasharp=false, PreFilter=3, truemotion=true, plane=4, chroma=true)
LSFmod(strength=200, defaults="slow")

[Attachment 61226 - Click to enlarge]

If you're still interested I will post from the building, and one near the end where camera and people are moving. You will see Team OS (mainly using smdegrain) using lock frame stabilization will have more details on roof tiles, building detail, but less noise on the building, windows. For the end shot, the wall detail "bricks" and text will be more clear on wall , less noisy, and there will be less ghosting for moving people. I can post a package with the stabilized frame set, output denoised, and output oversharpened (so you can compare with your oversharpened result, or use another sharpener) . You can improve over what you posted using Neat Video, but there are no tricks that I know with NV that can improve it beyond team OS (believe me I tried, I play for both teams) - the main limitation is the (relatively worse) temporal denoising quality and the TR limit of 5 for NV - it cannot take advantage of more stabilized frames. Believe me, I've done these tests before on 100's of sources and Neat since V2

Here are some cropped previews. The smdegrain+filters version is oversharpened to roughly match. The architecture shots are enlarged 2x. Notice the how much cleaner the smdegrain+filters version is for the walls, windows. The window border (the dots) are visible. The roof tiles, etc.. . The wall text and brick outlines are more clear, less noise. The people have less ghosting


[Attachment 61216 - Click to enlarge]

[Attachment 61217 - Click to enlarge]


[Attachment 61218 - Click to enlarge]

[Attachment 61219 - Click to enlarge]


[Attachment 61220 - Click to enlarge]

[Attachment 61221 - Click to enlarge]


[Attachment 61222 - Click to enlarge]

[Attachment 61223 - Click to enlarge]


[Attachment 61224 - Click to enlarge]

[Attachment 61225 - Click to enlarge]

[rrats]

The avisynth way does look better..

[lordsmurf]

You can't build a good comparable Still camera set under 3000-3500$

If you include a quality stepper lens or two, Atomos recorder, good mic ... sure, do not disagree here. But the system will easily surpass the lower-end older camcorder you've mentioned.

Even the audio elecronics is a scrap in still cameras, so it is not enough to buy a Good microphone, but you need external audio recorder. A comparable quality zoom lens system const 1500 - 2000 $.

The built-in mic is crap in almost anything.

Still cameras had not enough dedicated buttons for various functions, so their usage is more difficult.

I don't really see that camcorders are any better. Some mirrorless have moved to touchscreens now.

you always must buy 2 cameras/camcorders to make different angle shots. Two device is the minimum.

Why? You can shoot B-roll later. Or before. It really depends on the event or shoot. This isn't too different from a photo story detail shots.

And with Cam Link 4K, you can really shot lossless quality.

Are you sure about that? It's not on the supported list: https://www.elgato.com/en/cam-link/camera-check

Samsung S21 phone shot 8K video, which is better than most of still cameras which usually shot only in 4K....

What kind of "still camera" are you using?

Higher resolution means better quality

You're measurebating again. Resolution doesn't mean better quality whatsoever. The resolution of a digital device is merely the palette, not the resolve, the DR, etc. The sensor and optics determine that. Phones have crap sensors and optics. It's "good enough" to video singing Happy Birthday to junior, but it's not going to win you a video contest where technical quality is a consideration. You don't shoot TV spots or movies on phones -- but you do sometimes (often now?) shoot those on mirrorless systems.

and less halo effect. Halo effect is always there on every time for a some degree when you make photo or video which contains high contrast areas. (Just enough to make a magification to see it)

I'm glad to see this thread taught you something. You were clueless about halo/ringing until a few posts back. You're still wrong about some things (ie, "always there", it doesn't have to be), but at least you're learning.

[rrats]

but you do sometimes (often now?) shoot those on mirrorless systems.

lots of televisions these days use dslr/mirrorless cameras to shoot non-live television shows.

[Truthler]

You can't build a good comparable Still camera set under 3000-3500$

If you include a quality stepper lens or two, Atomos recorder, good mic ... sure, do not disagree here. But the system will easily surpass the lower-end older camcorder you've mentioned.

Even the audio elecronics is a scrap in still cameras, so it is not enough to buy a Good microphone, but you need external audio recorder. A comparable quality zoom lens system const 1500 - 2000 $.

The built-in mic is crap in almost anything.

Still cameras had not enough dedicated buttons for various functions, so their usage is more difficult.

I don't really see that camcorders are any better. Some mirrorless have moved to touchscreens now.

you always must buy 2 cameras/camcorders to make different angle shots. Two device is the minimum.

Why? You can shoot B-roll later. Or before. It really depends on the event or shoot. This isn't too different from a photo story detail shots.

And with Cam Link 4K, you can really shot lossless quality.

Are you sure about that? It's not on the supported list: https://www.elgato.com/en/cam-link/camera-check

Samsung S21 phone shot 8K video, which is better than most of still cameras which usually shot only in 4K....

What kind of "still camera" are you using?

Higher resolution means better quality

You're measurebating again. Resolution doesn't mean better quality whatsoever. The resolution of a digital device is merely the palette, not the resolve, the DR, etc. The sensor and optics determine that. Phones have crap sensors and optics. It's "good enough" to video singing Happy Birthday to junior, but it's not going to win you a video contest where technical quality is a consideration. You don't shoot TV spots or movies on phones -- but you do sometimes (often now?) shoot those on mirrorless systems.

and less halo effect. Halo effect is always there on every time for a some degree when you make photo or video which contains high contrast areas. (Just enough to make a magification to see it)

I'm glad to see this thread taught you something. You were clueless about halo/ringing until a few posts back. You're still wrong about some things (ie, "always there", it doesn't have to be), but at least you're learning.

Again, the touchscreen appeared on videocameras first, it appeared on still cameras later. Touch screen can not supersede dedicated buttons, because of the menu system, you must reach a lot of options after some touches, which is loss of time when you record quickly happening events.

"The built-in mic is crap in almost anything." Wrong. The built in mic of top consumer camcorders are on the same level like 500-1000 $ microphones. They are superior in indoor recording, especially for indoor VLOGs. Only noisy windy outdoor records can be a problem, when you have to use external microphones with "cat furry".

Who cares that Elgato did not test its Camlink for this camera, if it works?

I know about halo, when you were just an infant.

[Truthler]

It would be more interesting to post a response video with SM-Degrain (I guarantee that it will be worse), and you can write chriticism after that.

SMDdegrain alone? What happened to team open source? Are you limiting it now ?

Worse than what? I guarantee it won't look as bad as the mess you posted if you allow team open source to compete.

And how can you "guarantee" anything when you haven't used any of the filters properly (including Neat Video) ?

The way people use Neat and SMDegrain , just basically default settings, - in general I think Neat will produce better results and faster. But it's not the BEST results. Neat might be good for general user, but it does not produce the best results for power user. Neat is the editors choice because it produces good results, fast. Editor doesn' t have time to invest for the best results.



If you use SMDegrain + pre stabilize, pre/post filter, it will produce similar or better results in general to Neat Video. But much slower - in the order of 20-30x slower (!) for the denoising step itself. SMD will produce noticably better results on sections like the building, newspaper. Why? Because the TR for the avs+ version is not limited to 5 like neat video. For "regular" scenes with moving people/objects, that you prestabilize, the difference will not be as large (because you use a lower TR to reduce ghosting), but open source will still produce similar or better results with less ghosting. I mentioned it earlier, the temporal denoising is still a slight relative weakness for Neat V5. At large TR values, SMD approaches the quality of stacking methods, but SMD does not have the contamination issues with large peripheral borders, or as large issues with FG ghosting compared to mean/median stacking. Aligned areas or frames will still be the best with stacking methods using mean/median

Understand that I'm using "lock frame" stabilization - because it produces the best results for any denoiser using temporal denoising, including Neat Video. Each "n" +/-x frame is stabilized on that "poster" frame, the single output frame kept, the rest discarded. Repeat next frame. Ie. each current frame requires a set of stabilized frames in reference to that "poster" frame. You can automate some of the tasks with scripts and keyframes. This is why it's complicated and takes a long time. This way, you don't get peripheral border contamination for small to medium borders, even with Neat. But the question was about "best". This is the best method, even on scenes with moving objects (you can use slightly lower TR to reduce ghosting). And if you use everything properly and include complicated filter chains, it's difficult to compete with a full array of options of open source. This is the reason I don't post full video - it takes too long to get the absolute best - you'd only do it "for real" .

I didn't post the newspaper using smdegrain with large TR, but the results are clearly mean stack >> smdegrain > neat in terms of signal to noise ratio. (And there are no tricks I know with Neat Video that can improve it more)

AVISource("445-506_stab_ut_10bit422.avi")
SMDegrain (tr=20, thSAD=1200, thSADC = 800, refinemotion=true, mode=0, contrasharp=false, PreFilter=3, truemotion=true, plane=4, chroma=true)
LSFmod(strength=200, defaults="slow")

[Attachment 61226 - Click to enlarge]

If you're still interested I will post from the building, and one near the end where camera and people are moving. You will see Team OS (mainly using smdegrain) using lock frame stabilization will have more details on roof tiles, building detail, but less noise on the building, windows. For the end shot, the wall detail "bricks" and text will be more clear on wall , less noisy, and there will be less ghosting for moving people. I can post a package with the stabilized frame set, output denoised, and output oversharpened (so you can compare with your oversharpened result, or use another sharpener) . You can improve over what you posted using Neat Video, but there are no tricks that I know with NV that can improve it beyond team OS (believe me I tried, I play for both teams) - the main limitation is the (relatively worse) temporal denoising quality and the TR limit of 5 for NV - it cannot take advantage of more stabilized frames. Believe me, I've done these tests before on 100's of sources and Neat since V2

Here are some cropped previews. The smdegrain+filters version is oversharpened to roughly match. The architecture shots are enlarged 2x. Notice the how much cleaner the smdegrain+filters version is for the walls, windows. The window border (the dots) are visible. The roof tiles, etc.. . The wall text and brick outlines are more clear, less noise. The people have less ghosting

Dear Poisondeathray!

I miss your video.

[rrats]

"The built-in mic is crap in almost anything." Wrong. The built in mic of top consumer camcorders are on the same level like 500-1000 $ microphones.

Actually, those little video-blog cameras (aka ultra expensive pointnshoots) have best microphones, as they were made to operate without an external one

[Truthler]

"The built-in mic is crap in almost anything." Wrong. The built in mic of top consumer camcorders are on the same level like 500-1000 $ microphones.

Actually, those little video-blog cameras (aka ultra expensive pointnshoots) have best microphones, as they were made to operate without an external one

Even the ceapest camcorders from the well known brands have possibility to connect it to external 3.5mm jack microphone, and they have hotshoe connectors too, thus you can even buy a XLR audio box converter/adapter for them.



Yes, this hot-shoes also appeared firs on videocameras, similar to the turnable LCD screens and touch schreens. So the camcoirders were/are the most innovative tools, where new things appear.

[poisondeathray]

Dear Poisondeathray!

I miss your video.

I already explained , I don't have time to upload full processed video. I explained how you can improve neat video results or any temporal denoiser such as smdegrain TR = x>1 by pre stabilizing against each reference frame. Please re-read the post that you quoted, if you don't understand, ask specific questions.

Did you want me to upload an undenoised/unprocessed but stabilized frame set used for those frames in the screenshots along with the scripts used? (frame 1083,1683 - because those are the ones you chose for earlier comparison screenshots with selur) . I can post the mean stack version too - that's the best for the architecture

[rrats]

Even the ceapest camcorders from the well known brands have possibility to connect it to external 3.5mm jack

I highly doubt. All the camcorders I have have no 3,5mm jacks.

[Truthler]

Even the ceapest camcorders from the well known brands have possibility to connect it to external 3.5mm jack

I highly doubt. All the camcorders I have have no 3,5mm jacks.

Why did you buy videocameras from the bottom of the market?

Sony HDV-Z20, the cheapest Sony camera alsho had the shoe, where you can connect even XLR audio box, or other external mics.



Cheap Panasonic

[lordsmurf]

I highly doubt.

The more this troll posts, the more you realize he doesn't know what the hell he's rambling about. Everything from cameras to shooting to processing. Completely f-ing clueless from start to finish. It's like a Q-anon member for video. Most of his claims are divorced-from-reality videographical bullshit.

This clusterf'ck of a thread needs to be locked already.

Baldrick, bueller, Bueller?

At some point, you have to believe that he's just taking the piss, just screwing with people intentionally, ie being a true troll. A sad little person.

[Truthler]

I highly doubt.

The more this troll posts, the more you realize he doesn't know what the hell he's rambling about. Everything from cameras to shooting to processing. Completely f-ing clueless from start to finish. It's like a Q-anon member for video. Most of his claims are divorced-from-reality videographical bullshit.

This clusterf'ck of a thread needs to be locked already.

Baldrick, bueller, Bueller?

At some point, you have to believe that he's just taking the piss, just screwing with people intentionally, ie being a true troll. A sad little person.

Yes yes, these primitive personal attacks are clear sign of your lost debates...

[Baldrick]

I guess it's time to close this.