DVX100 Artifacts, Is this normal?
 

I'm experiencing extreme blockiness with footage shot with my DVX100. It appears during scenes with smooth continuous tone areas, like a shot of a wall lit with natural lighting. The colour blocks are huge! and they flicker randomly. I'm not sure if all DVX100s exhibit this problem because the footage is unacceptable for me, after it is rendered out of AE with colour correction or any type of image manipulation. I have not experienced this type of anomaly with footage shot with any other camera. It is hard to explain without showing samples. I built a web page that illustrates what I am talking about. Please take a look.

http://www.pigoinky.com/dvx100/

I encourage anyone who owns a DVX100 to run similar tests to see if you have the same issues. The final quality, for me, is totally unacceptable and I am seriously considering getting a refund on this camera. I shot an interview segment and noticed it during playback, where the interviewee was lit from a window and her surrounding was in 10% shade. The shaded areas exhibited an extreme amount of this "macroblocking" that was noticeable from the first generation. I haven't done exhaustive tests, but the tests so far have shown the DVX100 handling of smooth gradients to be abysmal.

Btw, I tried contacting Panasonic Technical support, but was unable to get through. I've heard horror stories of people trying to get technical support over the phone, so I'm bringing this issue to dvx100 members in hopes of figuring out the real issues to what I have been seeing.

Judging from your example frames, I would want to diagnose this problem as contouring or gradient banding, which is common in shots of this type from prosumer digital video camcorders. The XL1 suffered from the same effect under certain conditions... see a three-part article on my website:

http://www.dvinfo.net/canon/articles/article10.php
http://www.dvinfo.net/canon/articles/article5.php
http://www.dvinfo.net/canon/articles/article18.php

If the DVX100 is implementing an 8-bit DSP, then it wouldn't surprise me if this is indeed a gradient banding issue. There is no fix; instead, one needs to take care not to shoot smooth continuous tone areas. For what it's worth, even high-end broadcast cameras can suffer from this, so it's not at all uncommon.

Chris,
Thanks for the great info on contouring. If it's not one thing (audio synch with the DVX100) then its another (contouring with just about everyone).

Trying to make a final decision on which digital camcorder under $25,000 to get is like running a maze. For some of us, it's worse than buying a car!

BTW, you are one lucky person to be living in San Marcos. That place has got to be one of the best kept secrets in the U.S. Small university town, coeds, clear water river running through it (Frio? Guadalupe?), and easy access to Austin and San Antonio. Did I mention the Hill Country just minutes away? Beautiful, just plain beautiful.

Thanks again for the great info. Nick

Thanks Nick, the river is the San Marcos River; it's headwaters come from a natural spring near the university. It's one of the cleanest rivers in the United States. Next time you watch Sam Peckinpah's "The Getaway," you'll see Steve McQueen jumping into it. San Marcos is a wonderful town still, despite the outrageous growth in the past decade or so. Where are you?

I've had the DVX for over 2 months now and haven't noticed this problem. I've also not done the kind of extreme color correction you're doing, but I'd be curious if other cams do the same thing.

Chris, I've lived in Austin since 1989 and have never visited San Marcos. Anything worth seeing on a day trip? Did you end up getting a DVX? Or still working with a Canon?

Skip Hunt

I just did a similiar test, and while there is banding (as should be expected from an 8 bit source), I have no macro blocking.

Normal
http://home.inreach.com/bobinick/1normal.jpg

Equalize
http://home.inreach.com/bobinick/equalize.jpg

My test
 

Here's my test:

http://www.celout.com/dvx/

I can see the normal effects of DV compression, but nothing as extreme as Poojja's test.

The banding and possible macro blocking will be more pronounced with a less colorful image . If you look at Poojja's images you'll see it has a fairly even shade of grey. Your choice of a more mixed color background will hide contouring, or banding.

Jeff

Re: My test
 

OK I think there is something seriously wrong with my DVX100!! Your pics look so nice and they are so smooth, even telling are the equalized images. You have no extreme form of macroblocks that mine have. I re-did the same test, this time using a blue gelled light from a 10watt halogen. And sure enough, even with coloured lights, the banding is still there. I don't think this is banding, because the same scene shot with a GL-1 has none of those ugly macroblocks.

I re-did my test and have updated the web page with more pics.

http://www.pigoinky.com/dvx100/

I think there is a problem! :(


<<<-- Originally posted by Joe Kowalski : Here's my test:

http://www.celout.com/dvx/

I can see the normal effects of DV compression, but nothing as extreme as Poojja's test. -->>>

Poojja,

Your comparisons are inconclusive because they are shot under different conditions, and are not even the same subject. If you want to do an accurate comparison then you must test cameras under identical conditions. Barring all that I do believe you have a problem with the blocking and all. Some banding is expected in an 8-bit device. Yours, however, seems to go beyond normal expectations.

If you want to do a valid comparison, I would suggest that other users shoot the dark blue portion of the sky. Barring some atmospheric conditions and pollution, everyones sky is lit pretty much the same. Then compare results. Random lighting and subjects will produce random results of little use.

Jeff

Chris,
My wife and I currently live in Dickinson, Texas (between Houston and Galveston). My wife is from Austin and we lived there for the better part of 15 years during which time we attended and graduated from U.T. (liked it so much I graduated twice).

Yea, the San Marcos River. I was thinking of the river that runs through New Braunfels..... man, is that one ever cold!

Love this forum, thanks and Hook 'Em Horns. Nick

Hi Nick -- lots of Texans on this board; some are displaced in other parts of the world... one big happy family here.

Skip -- the best time to visit San Marcos is when the University is in session; swing through Sewell Park by the river on a sunny day and you'll see what this town is all about.

Poojja, Can you test again without cine gamma?

<<<-- Originally posted by Andre De Clercq : Poojja, Can you test again without cine gamma? -->>>

I too would like to see you repeat this test without cinegamma. Cinegamma yields some very interesting and nice results, but there seems to be a bit of strangeness that goes on with cinegamma selected under certain shooting conditions, particularly low-light.

BTW: where's Michael Pappas been hiding? He was all over the place joining the estatic DVX minions, but haven't seen him around the boards in weeks. Perhaps he's happily shooting? I shot some face closeups yesterday with the "THIN" vertical res setting, cam setup "0" IRE instead of 7.5, detail at "0", Cinegamma, 30p, etc. then color corrected, cropped and stretched in AfterEffects 5.5PB to widescreen 16:9. Then encoded to mpeg2 with high bitrate and quality, burned to DVD, projected from prog-scan DVD player via component cable to XGA LCD projector set to 16:9. I also did everything the same to the source footage except the cropping and stretching to 16:9 for comparison. The results were VERY good. The widescreen stuff looked fantastic. The only degradation I could see was a very slight softening from the crop and stretch. But, because I shot with the vertical detail on "THIN" and the detail set up to "0" from my usual preferred "-2", it basically balanced out and looked very, very good. I'd urge those of you with the same tools and a prog-scan player with large screen display, plasma, projector, etc. to try the same thing and let me know what you think.

Skip

First thing Monday, I am going to return the DVX100 to my dealer to get new replacement. Joe K's test of his wall shot really did point out the problems I am having with my camera. Had I shot the exact same footage he did, I am pretty sure I would be getting the macroblocks. The problem is that the huge colour chunks occur anywhere there is a areas of the same colour, which means about 95% of my footage is affected.

Again, unacceptable. I haven't tried it with cinegamma off, but what would be the point in testing it. If it doesn't show artifacts with Cinegamma turned off, I still would not be happy NOT being able to use the cinegamma functions.

Panasonic should really look into this, because I am quite pissed with them right about now. All the footage that I have shot so far with this camera is flawed, and I don't know how I am going to hide these artifacts once I start editing it.

All gamma manipulation needs ("eats") grayscale resolution. If there is not enough remaining bit depth before the DCT tranforms happens those color and grayscale clustering effects can happen (what you call macroblocks). That why pro cams use up to 16bit preprocessing before the 8 bit quantization for DV happens...

I was under the impression that Cinegamma only affects the grayscale values. And DV 4:1:1 retains full resolution in the luminance channel. But the problems are occuring only with colours.

My test also included bringing a frame grab into Photoshop and converting it into the Lab colour mode. Looking at the individual channels revealed no macroblocking in the luminance channel but extreme blockiness in both the a and b colour channels. Footage from other camera sources did not have the same extent of the artifacts.

The fact remains that the colour channels do exhibit these artifacts and may be a result of the cinematrix settings. My GL-1 and Elura 40 do not exhibit these artifacts so it may have to do with the cinematrix settings. There wouldn't be a point comparing it against my DSR500 either. I can only assume there is a problem with the DVX100, as compared to my lower end cameras which do not show these artifacts. Having these huge artifacts seems counterproductive, if colour fidelity is hugely compromised as a result.

When I get a replacement camera, I'll post my findings.

Gamma correction is done on the RGB signals before YUV is being "made". To0 low bit/pixel count before gamma correction and rough (DC) Q values after DCT transformation can result in your the problems you get.

You are probably right. Histograms in PS of my frame grabs show gaps. I never thought that to be right.

What, then, would explain why my DVX100 suffers from severe pixellations while frame grabs from other DVX100 owners seems to be fine? (and fine even from exact same tests made with my GL-1) I think my unit is faulty. The reason I posted is that all my footage exhibits these blocks, not just the rare instance. I've worked with DV material for several years, and it is the first time I have come across something like this.

To be fair, as Jeff pointed out, the lighting conditions of my test may have been too high contrast to bring out the problem. I will try something more subtle if I have time.

Pooja - if you do exchange your camera for a new one, let us know if it exhibits the same problem. I am hoping this is not something that affects all DVXs and I just haven't noticed yet.

Update... five weeks later...

The problem is serious enough that both my dealer and Panasonic have offered to refund my money or keep a new camera, although they acknowledged the problem does exist and I have a choice of getting my money back.

Considering the problems I've been having with my camera I've decided to get a refund with a possibility of purchasing same camera model a year from now when I can be certain that all the bugs have worked out of this camera. I am very disappointed and all the footage that I have shot up to this point is fatally flawed.

Pooj

ps: thanks to all who have submitted screen tests to isolate this problem. Panasonic, although slow in identifying and rectifying this issue, they have been very gracious and considerate of the problems I've been experiencing.

Poojja -
sorry that things didn't work out. If you decide to re-buy the camera, you may want to consider a dealer such as B&H that seems to be a bit more receptive to problems like that when exchanging/refunding. Luckily, this seems to be an isolated incident... hope you have better luck next time around.

Joe,

I returned my original camera to Panasonic. Once they found the problem, I suggested I get a replacement camera from my dealer, which I did. The problem was found on the entire batch of cameras at the dealer. Starting with serial numbers J2Txxxxxx. I was shocked. Incidentally all the cameras that the dealer had were sent back to Panasonic. Just yesterday I received word that the new batch supplied to the dealer from Panasonic exhibited the same problem but to a lesser degree than was exhibited on my original camera.

For me, that was unacceptable. They offered me the new camera but the problem was still there but to a lesser degree. (Panasonic tech confirmed over the phone)

So I finally decided to get my money back fearing that over time the problem would get worse, which would be not good.

The problem does affect more than just my camera.

Am I being too nitpicky with image quality from the DVX100? No, because my GL1, XL1, PD150 and my DSR500 do not exhibit these problems. The dealer said it may be something to do with the 24P interpolation, but I fear it's not the case.

Pooj

I've been having some strange discoloration problems as well. I've got a fresh thread on the subject and am hoping it can be reasonably explained. However, I checked my serial number and it too starts with J2TD... J2TD00752 to be exact. Does this mean mine is definately flawed? Which vendor did you deal with?

Skip Hunt

Hmmm,

I think I might have discoverd an interesting development with this thread. Pooja is having these macroblocking problems with a capture via firewire from FCP using the AppleDV Codec. The other two examples on this thread that do not show the severe macroblocking, just simple banding artifacts were captured from Vegas Video which I believe has it's own special DV Codec from Sonic Foundry as of version 3. On www.24p.com there's a comparsion between frame grabs from FCP and DVXpress, with the XPress framegrab being much smoother. These macroblocking artifacts are only in the chroma channel, and DVXPress smooths the chroma channel while AppleDV does absolutely no smoothing, it just simply repeats the color information four pixels over. I use FCP and have experienced similar macroblocking in dark areas of the frame, even though I cannot isolate these artifacts on an NTSC monitor when being played out from a deck or the camera, either via SDI or Y/C (from the camera, SDI from a DVCam deck, and although I have another thread about problems via SDI, they are not the macroblocks that Pooja is reffering to). So, is this problem a characteristic of the AppleDV codec decoding the chroma information from the DVX100? Do other cameras (PD150, GL1, XL1s) perform some sort of chroma low-pass filtering before recording to tape to elimate the 32x8 pixel chroma DCT blocks that the Panasonic does not do, and hence the AppleDV codec is showing problems since it too does not do any chroma filtering? Just something to think about.

Jason

I have been able to find one, maybe two instances of this artifact happening in five hours of footage. And when it does happen, it needs Photoshop Equalize to become really noticable.

Skip, I know we got our cams at the same time, and mine is serial J2TD00561, maybe it just depends on the camera that you get and not what updates Panny has done. I would assume mine is #561 off the production line.

So does anyone have a model earlier than this or around the same time, and what are your results?