CHROMA NOISE in HD10s
 

Yellow-green blocks are dancing on the skin or same-color surfaces like walls, when shot with JVC HD cams.

See examples here:

http://fancyflix.com/hd10.htm

I had nothing like that with my previous (non-HD) camera.

Is this a result of 1 CCD?

Is my camera malfunctioning?

Please opine based on your own experience with these HD cameras!

Thanks :)

Is it possible all this chroma noise has more to due with the higher resolution than with the one ccd? Or perhaps the higher resolution combined with the MPEG compression.

Not currently owning this camera, the resolution in your shots looks amazing.

Does anyone have similar images comparing the edge enhancement between the two models?

Brian

I'm repeating this post from another thread as it asks the same question.


First of all you need to determine, is the chroma noise I see ALL in my orignal footage or is it being enhanced by my playback equipment?
While there is most definately significant chroma noise in the original footage shot on the HD10U, I have found that the software and hardware you play back through may enhance that chroma noise a great deal and give a very distorted impression of just how much chroma noise there actually is.
Worst are the software playback players like Elecard and VLAN. Next is NOT playing back on an HDTV (not a computer monitor) that is capable of 720p native format. There are very few HDTV's that are actually capable of this, most do upconverts to 1o80i or downconverts to 480i, on the fly, so I would say that MOST people are probably getting quite a distorted impression of how much chroma noise is actually there.
Though I am by no means reccomending a Monivision monitor to anyone, the amount of chroma noise I see in flat areas of color (where it is most obvious) is far less when viewed directly from the camera, or after transfer to D_VHS, on this 720p capable monitor.

I hope this helps.

While I do not own that camera, I do know that 1CCD cameras have MUCH higher Chroma Noise than a 3CCD camera. I can't wait until they make a 3CCD HDV camera. It will be awesome.

My experience with the camera shows chroma noise is much as experienced with miniDV. Paul is also right about the display problem, be sure of your monitoring gear. Head to head, the DVX100, XL1s and HD10 displayed pretty much the same amount of chroma noise in the same shooting environnement, not exactly but even the miniDVs where not 100% equivalent. We have to understand that these videos are compressed by a very high though equivalent factor, miniDV (NTSC) is 4:1:1 (please don't tell me there is no chroma noise with 4:1:1 compression, I never shot with a miniDV camera that did not produce some chroma artefacts) and miniDV (Pal) and the HD10-HD1 use a 4:2:0 compression. Both mean that the chroma is reduced to 1 pixel out of 4 of the original image saturation, the 4:1:1 process is linear (on the same line thus a rectangle of 1X4) and 4:2:0 is bi-directional (horizontal and vertical for a square of 2X2, of course not very good in interlaced material but correct in progressive). Degradation in the chromatics (both red and blue) are equivalent in both types of compressions (of course the green channel is derived from the red and blue ones). A thourough examination of miniDV and HDV showed to us that high contrast rations (night clubs, night scenes, etc) are much better handled by the HD10 because of the fact that the compressor diminishes chroma in low light areas, thus significantly reducing chroma artefacting and improving the blacks over DV. The one CCD versus 3CCD thing here is very difficult to determine since this one CCD camera does not produce color in the same way as traditional 1CCDs. I suggest you read Steve Mullen's article on that subject following this link:

http://videosystems.primediabusiness.com/ar/video_ccd_counting_needed/index.htm

But let's be realistic, you have to be very efficient and to know the HD10 well to harness it's full potential. You need to understand light in a videographic way and to be open minded but some sequences (posted here) show very good examples of this camera's strenghts and weaknesses. Using the camera in full auto mode will give you worse results that auto mode with the VX2000, XL1s, DVX100 or PD150, in fact, in most cases it will give you crap.

Be open minded and this camera will give you outstanding results.

Eric,

Thanks, that was a very useful and to the point summarization.

Troy

Excellent post Eric.
The artifacts that are being spoken of I see with all video played back on my pc. I would add a salmon pink to the other two colours described. The more compressed the more the weird colours. It seems to be an element of software decoders and enhanced I find on ATi and NVidia cards. I also would like to know if you were using a software decoder Alex.
Ken

Ken, the color noise looks the same when played directly from the camera's Component ouput to HD monitor, or when viewed on PC via software player after the capture.

So this effect is *not* related to any PC-based software processing, but comes directly from the camera.

You wrote:
"I also would like to know if you were using a software decoder "

Please refer to the page in my original post - it describes in great detail how the snapshots were made.

"The artifacts that are being spoken of I see with all video played back on my pc"

I don't. My previous SD miniDV Sony VX2000 did not produce anything close to this chroma noise, and I pulled quite good blue screen works out of it with good lighting.

With HD10, chroma noise is so great, forget about the blue screen.

Also processing: try increasing saturation even slightly. You'll see weird lepro stains all over the place; the picture is falling apart instantly.

It was not so with VX2000.

Thus my question to people who *actually own/use* HD10: do you have the same issues or is my camera not adjusted properly?

"With HD10, chroma noise is so great, forget about the blue screen."
Well you have my attention. Even if you calculate in the increase of resolution?
My experience when getting that type of "noise" ,when using DV format is, when I gain-up in low light, or under expose in general. Or re-render to often.
The question is as follows:
Is the "croma noise" more or less than DV, given the increase in resolution? We have to talk apples to apples here. If we zoomed in the JVC-HD to SD resolution, what would be the croma noise diferance between itself and a quality DV cam?
Ken

I believe the color noise is in the camera not software decoders, crt tv'v have a little softer image than the pc so it tends to look better.

To compare this to DV color noise would have to be done by some level rating and compared percentage wise to the HD. With any footage it looks good on smaller levels. Watch a VHS tapes on a 9 inch tv it looks great, but put it on the 60 inch. big screen and whoaaaa crapola. The color might be better than dv but the size of the pic allows you to see more Flaws.

I agree, the chroma is way to poor for blue screen (unless you have some stellar equip. and skillz :-) I have shot EXT. night, EXT. day, flouresent, and stage spotlight. I'll post some frame grabs soon. After tweaking it a bit I would like some other opinions on if's better or not.

-S

The color shifts are probably not a compression thing, most likely they are related to the "automated de-saturation process", not as a consequence but I think that this particular chip seems to be much less chroma accurate in low light conditions, particularly on the skin tones. Witch is not the case of the sony models of miniDV I agree. Still with decent lighting it does real good, even in green screen, I suspect blue to be equally good.

Looking on the footage on a HD monitor gives real credit to it and surprisingly, those color shifts are much less apparent. Alex is right though, tweeking the image will worsen that in post, a good reason to increase the light. This camera does very good in fair lighting conditions.

Still, there are conditions where the image is totally screwed up. This camera is, needless to say, less than perfect but it does green screen without problem on my part. The increase in definition really is a major issue here. Let's be optimistic and assume that the next generation of HDV will have these issues solved :)

As for the chroma noise Alex showed on his site, it is indeed a HD1-10 related kind of noise. It appears in areas that are not enough lighted and in the red-orange-brown-pink areas of the image mostly and in low contrast zones and/or situations. That is why I think it is related to the de-saturation process. The techs might have skipped this problem by providing a wise de-saturation protocol to avoid the weird color shifts. It is not likely to be a compression noise. I thought that the de-saturation was to keep the bitrate but it might be related to that instead.

Still the chroma noise created by compression is somewhat equivalent to DV. But this noise is not the same. The combination of both indeed destroys a good image so we have to be cautious.

Based upon the easy compression of green and the difficult compression of blue and red, plus the likely greater sensitivity to green -- I expect Eric is correct about green-screen working fine.

Also, the chroma noise projected to 6-feet or on a $4000 JVC HD monitor is not terribly high. I think Paul may be correct that decoders vary in their quality. (Paul did you get a replacement Monovision?)

Yes, it is higher than modern 3-chip DV camcorders, but it is not all that terrible.

I remember thinking it was a bit like the VX700 which did not have a chroma noise filter. The next gen DV camcorders had a chroma noise filters and the noise went away. But at a price.

DSP filters trade resolution for better S/N. I'm not sure if that's what we want.

New info on HD10's chroma noise.

If you shoot using cam's B&W effect, the footage is originally black and white. There's no visible artifacts at all in the ares where color footage does show chroma noise.

If you shoot in color, then you have color noise.

I expected cam's B&W effect footage to be superior to the color footage *de-saturated in post*.

After all, if the noise is already there, and you simply take away all color, you still should see some brightness fluctuations where chroma noise "stains" used to be, right?

Wrong! I was surprised to discover that software desaturation works *equally* well as the cam's own B&W effect.

There's absolutely *no* brightness spots in place of color noise spots after desaturation.

Which means that the brightness of the "noisy" spots was perfect, just the color shifted.

Anyone cares to analyze this?

My only comfort at this time is that yes, I can acquire color footage and then desaturate it in post at will - without any loss of quality or any extra luma noise.
By the way, 50% desaturation already takes care of the chroma noise spots - but of course, almost all the "normal" color is gone as well...

The reason why it works equally well is simple, the shifts in the chroma are color shifts, not intensity shifts and on normal conditions, a desaturation post process will give results as good as the BW filter on camera. A lot of colors produce the same grey level if they display the same lightness or almost identical.

But in low light high noise situations, it is much better to use the BW filter since the de-saturation (unless made in YUV or lab by discarding the two chroma channels) will produce visible blocks in the final desaturated images since the chroma shifts are combined with the low light chroma compression noise.

De-saturating the image a little indeed does the trick of lowering the noise but at some expense...

Steve, I should probably report that the Monovision has been a nightmare for me. Though the picture is great when it works, it has been totally unreliable in my case. I am now awaiting a refund on mine, as I'm on my 3rd failed replacement set and have had several repair visits to no avail, I may have just been unlucky, but that's the sorry situation. If it was reliable I would probaly keep it, Monivision service has been very good about repairs and replacements though. It also has issues with the black levels, which are not at all consistant in dark scenes, and also with the VisionBox converter, which produces green vertical lines on screen. Not a good situation. If you know of any other native 720p capabel monitor I'd really like to know. I'm also interested in that $999 HD projector from B&H you mentioned, can you tell us the model?

Thanks

Paul, I'm really sorry to hear about your monitor.

The X1 from InFous -- i think.

But for $1300 you can get the Sanyo Z1 which I have and recommend highly.

The X1 uses a spinning color filter which cause some/most folks to see rainbows on moving objects.

Head to Video Systems where I review the Z1.


The color shifts are Color PHASE errors (PM) while brightness changes are Color AMPLITUDE AM) errors.

But, while I see PM, I also see AM on saturated red and blue areas.

AM is far more annoying.

I agree with Steve that Amplitude modulation errors are more noticeable. I did see a lot of yellowish skintone shifts on HD monitors and it looked just fine to me, it is not all that bad

Does anybody else have HD10's footage snapshots?

If yes, please post the web links so we can see how chroma noise propagates itself in YOUR camera.

Mine are at:

http://fancyflix.com/hd10.htm

You know, I've looked at the footage off this camera on about 15 different monitors. Everything from LCD to progessional $20K HD monitors from Sony.

the level of quality of the monitor is a big difference and in my opinion, a CRT monitor is superior to an LCD.

If you're buying a new monitor for this camera, consider tube instead of LCD. The chroma noise is reduced, so in my opinion it's the monitor and not the camera in many cases (thought not all - traffic lights are trouble)

Cheers

I think if the footage is imported into the computer then a snapshot is taken and chroma noise is evident, iit s proof that it has come from the camera. Does this make sense? I mean the monitor is out of the question because it is not subjective viewing, but a frame from raw captured footage.

The questions that remain are:
Is the croma noise dependant on lighting levels?
Do the combinations of ample lighting and filters eliminate chroma noise ?
Is Alex's camera pooched?

Ken

These speculations on the origins of the high chroma noise fail to mention the extremely high noise-to-signal ratio that naturally results from dividing a single 1/3-inch chip into so many tiny sensors. It's the laws of physics, people.

A redesign could ameliorate this by using a much bigger chip or perhaps adding extremely sophisticated error-correction techniques in the chip.

But that would require HDV licensees to reconsider whether they want to continue treating this technology like Video **HOME** System for the 21st century.

(I hope they do just that.)

Hasn't it occured to you how much HD10 chroma noise looks like VHS noise?

We all know the camera could be better in many respects but that is not the issue.
We have to work around these limitations, or the problem we are having now, defining them.
I strongly believe that the chroma noise is proportional to light levels. If you look at Alex's pictures you can see the first two pictures have the most chroma noise, but they also are lacking in lighting. The out-door shot has higher levels of light but it is very direct giving us over exposed and under exposed areas. The shadow areas have the chroma noise.
It is my feeling that in studio conditions with pro lighting, chroma noise would be minimal. This cam seems like it can really benefit from a general high lumen environment. Almost treat it like a film camera. Shoot extreemly well lit, then drop the levels in post.
Ken

Ken is absolutely right, as I said earlier in this thread, with enough lighting it becomes less aparent, even totally absent.

<<<-- Originally posted by Ken Hodson : ...the monitor is out of the question because it is not subjective viewing, but a frame from raw captured footage.>>>


Obviously.



<<<Is the croma noise dependant on lighting levels? >>>


All 3 videos were made at different light levels from radically different light sources, yet chroma noise is present very visibly in each snapshot.

Were the chroma noise levels the same through all 3 shots?
The chroma looks to be less on the final shot, and is the only shot approaching adequate lighting.
Ken

As far as I can tell the chroma noise gets more apparent in badly lit situations. I will post some snap shots as soon as I receive my copy of Panther to show how good and how bad the image can be. In a perfect lighting condition this camera takes AMAZING footage. So the morale of the story is, invest in a good gaffer before shooting with this camera.

Sounds good Frederic. That will help us get to the bottom of this.
Ken

Stills from HD10U footage
 

Here are some stills from footage I shot over the last couple of days. This is all in natural light. Anything shot in manual mode by tweeking the settings comes out beautiful, especially when it's well lit (i.e. butterfly shots). I notice plenty of chroma noise when i shot in shadows, or down on the sand while it was heavily overcast. You will notice my little notes under each shot. Compare the benches. One was shot in manual and the other in automatic mode. Automatic mode is no good. I know that the colors aren't ideal but it's nothing color correction can't enhance. (I'm a color correction freak).

My favorite however is B&W, looks like super 16mm. Take a look and let me know what you think. Oh, one last thing...I just got my copy of Panther today and i still haven't figured out how to convert m2t to something I can open in quicktime so these are frame grabs from OS X and they are not at full resolution because my G4 monitor doesn't have full HD rez. The max is 1152 X 768.

However, it's pretty cool to know that we're looking at these huge shots and they're not even full rez. DV NTSC is only 720 X 480!!!

Here's the link:

www.lumieremedia.com/hdv

pathfinder.jpg:

the house's roof, doors, and back of the estate sign all have green stains - the same chroma noise I observed with my camera.

So what's your conclusion - we just live with it?

Alex,

It's night time now and I'll check tomorrow morning but I'm pretty sure these stains are on the paint of the building. I'll check and let you guys know.

If indeed these are stains from the camera it's a major bummer...however, the stains did not show up on the bright scene with the butterfly.

The best test would be to light a color pattern the best we can.

I will let you know what i find about the house.

Hey, is the only way to get m2t converted to something usable by FCP to by Steve Mullen's $100 4HDV solution?

Frederic, Mac solutions: sorry, can't help you here - I work exclusively on Wintel platform...

Butterfly pics indeed did not have any visible chroma noise as far as I can tell... BUT your image of the side of the moving car did.

It's alongside the whole side of the car, again looks like dark green blocks in random pattern.

I disagree that these artifacts are only showing in badly lit situations... just look at *my* pics - only the top one was badly lit, but the chroma noise propagated itself in all 3...

Also try shooting anything that is intensely red - say, building's canape or a car. You'll see some very lively chroma noise block movement where there supposed to be just a solid color... looks equally bad on both stationary or moving objects...

These green/yellow chroma shifts are a problem related directly to the camera. As it was said over and over again it happens on scenes with insuficient light mostly. As to Alex's question, yes, it is a problem but it is not that apparent (even unnoticeable in 95% cases unless you freeze the image) in moving pictures on a true HD screen, even downconverted. So yes, we should live with it but I do not see that much of a problem with it in good lighting situations. I think you might be nitpicking a little here, it is not so problematic when you see the image moving, it looks nice (I am difficult in terms of quality, I guess the shots I posted on DVI are a good proof of that) but looking at your shots on computer is not as looking at them on a monitor or on a projector. I do a lot of color correcting/tweeking and it is easily correctable by selective level adjustment, of course bad lighting gives bad results... DV, to me is equally hard to work with because of the compression noise but it lacks definition.

The point is these artefacts exist in some specific areas (brown, orange, pink mostly) in unsufficiently lit surfaces. You either live with it, correct it or else this camera might not live up to your expectations. But looking at still images is not a good reference for a motion medium. We intend to produce moving images is it not? So the point would be to try to see those artefacts at 30p, ideally on a decent monitor. Anything else is unconclusive.

Just my 2 cents.

<<<-- Originally posted by Eric Bilodeau : But looking at still images is not a good reference for a motion medium. We intend to produce moving images is it not? So the point would be to try to see those artefacts at 30p, ideally on a decent monitor. Anything else is unconclusive.

Just my 2 cents. -->>>

A valuable 2 cents. Anyone who has looked at a still of any motion medium can be shocked by what they see. Motion makes somethings go away.

When I was working with the DVX100 folks kept posting stills to compare to stills from something else. Drove me crazy.

Equally missing the point, IMHO, are those comparing the HD10 to a DV camera on a 27" NTSC monitor. Super 8 film looks great projected to a 1-foot screen. But try it on a 20-foot screen.

I absolutely agree Steve.

The deal here is if you can capture a great image with the HD10U with some effort (lighting) and I think the butterfly stills I posted prove that you can (although there was minimal effort involved) this camera is without a doubt a better solution than DV. Simply because of the resolution. Even if you stay in NTSC!

720 X 480 DV simply doesn't compare with 1280 X 720.

People have been chasing the cinematic feel since DV came out by adding grain and scratches, taking out frames, progressive scan, saturating colors, adding letter box, short of adding 'chroma noise'... In my opinion, nothing beats having true 16:9 and high rez.

Everytime I open that LCD and shoot in HD I get that cinema feel just by looking at it.

Just an artist' point of view with modest engineering knowledge.

"I disagree that these artifacts are only showing in badly lit situations... just look at *my* pics - only the top one was badly lit, but the chroma noise propagated itself in all 3..."

I would say all 3 pics lack in proper lighting. There needs to be more fill lighting. To the same level as when you see a film set and you can't believe how much light they have. That level of lumens I believe is what is needed for minimal chroma noise.

A good test would be an indoor shot, with the same subject in the same setting. No change. Then provide 3 distinct levels of lighting. Exposure levels should be noted. I think that is probably the only way for any conclusive comparison.
Ken

No offense Ken but the three shots you showed us are badly lit.

Hey, did you have the camera on battery or power supply when you shot these?

Sorry Ken, just realized that Alex showed us the pics.

I agree with what you said.

I'd like to bring the discussion back on the productive track.

Namely:

- *Moving* images actually make this chroma noise *more*, and not less apparent, to my eye. Therefore posting *stills* makes much sense - it is not a pointless disassembling of the otherwise good image, but rather bringing the problem into focus so to ask why is this happening.

- There's a point raised here that chroma noise only propagates itself when scene is "badly lit". It is further suggested that the "properly lit" scene should have as much light as Hollywood productions. To put it mildly, this point seems to be unsubstantiated to me. If you think otherwise, prove it with your own images.
Also, 2 out of 3 images I posted had a lot of light directed onto the subject, with one being bright sunlight, and yet all show almost the same level of chroma noise.

Specifically: have a human face with the beard "properly lit" or whatever, post the pictures that show the difference your approach made in reducing chroma noise (before/after proper lighting was applied) and this will make your point.

Same with the butterfly image posted by Frederic (who is so far THE only person to even post any images of his own): show the "bad lighting" example where the noise shows up, and then the "after" image with no/reduced noise. THIS will help. A single picture will not (I also have some shots where noise is not apparent - and? What is the circumstance under which the chroma noise goes away? This is not clear to me, please illustrate convincingly if you have a proved point).

- If someone is to post a reply, I'd ask this person as much as have used the HD10 cam in actuality, and/or have actually looked at my posted pictures.

Because if they had had, then:

a) they'd know that the camera does NOT display its exposure settings... neither during, nor after the shoot... so you can't tell what was the camera-chosen aperture with your setting of shutter speed in shutter-priority mode;
b) my image page tells you about it.

http://fancyflix.com/hd10.htm

Thank you for posting your own images that show how did you reduce chroma noise in your own setting.

Amen Alex! on all your points, especially the camera's inability to display exposure settings.

Hey, did you have the camera on battery or power supply when you shot the examples?

The reason I ask is that I just realized that my camera came with many noise reducers for cabling!? and I never thought of using them while shooting (DC powered) or while capturing.

P.S. I felt I had the right to post a reply since I've already posted pics...;)

"There's a point raised here that chroma noise only propagates itself when scene is "badly lit""
"To put it mildly, this point seems to be unsubstantiated to me. If you think otherwise, prove it with your own images."

True this is unsubstantiated, but isn't that the point of the thread to get to the bottom of the chroma issue? It is clear to see Alex that your 3 snaps are lacking in lighting. The best of the 3 is your outside shot that uses only one light source (the sun) at what appears to be late afternoon. I would say this is very unsubstantiated as to your statement that this camera HAS chroma noise in well lit environments.
But to give credit where credit is due, you Alex are the only one who has posted any pics that at least try to demonstrate the chroma noise. (Thanks for your shots too Frederic, but they don't help resolve the issue)
I think we are all on the same side here. We all just want to know the exact limits of this camera.
Ken