I've received an email from him, the gist of which I will not discuss here, but he did outline for me the steps that he took to create those stills. I am not disclosing the contents of a private email here, because he has also provided the same information about his capture method on a public page at his web site.

No mention is made of pulldown removal or deinterlace technique. Perhaps this will shed a clue as to why his particular samples directly conflict with the overwhelming amount of material provided by so many other people:

He says that he used this method to capture all of the stills from all of the cameras. He wants to know why, if his capture method is so bad, do some of the stills look worse than the others, and concludes that the reason must be the difference between footage. In order to bring this back to an orderly technical discussion, does anyone have any comments about this capture methodology in particular?

Wasn't there a claim that it was a denial of service? I suppose that might be true if you tried to run a web server on an Amiga or a PDA. A denial of service would either be enough traffic to crash your web server, or clog your internet service with requests such that you can't access the server or the internet.

Yes Chris, to be blunt, his methodology doesn't work and he should have seen this from the get go. It should have been painfully obvious to him that his finished product didn't begin to look as sharp as the original that was posted by Wes or the m2t from which it came. I'm not in the habit of making stills from any of my videos since I use a dedicated digital camera for that purpose. I posted a still from existing video to merely show him that what he posted was not indicative of the quality of the camera. In fact, I told him it's absurd to judge the quality of any video camera by the quality of the stills it produces or frame grabs for that matter. But as could be clearly seen from the still I posted, my pausing the video in-camera, hitting the photo button to capture, was an obviously 'purer' method to capturing more of the quality. I had no need to 'manipulate' anything in Photoshop other than to use it as a tool for downsizing to meet the 'attachment' criteria for AVS.

The ongoing irony here is that there are tons of still captures from the HV20 right here on dvinfo. Almost all of these are of far higher quality than what he posted. Wes' shots are just one example, there are others. This too should have been a 'clue' that something was not right with his 'methodology'.

Right, Ken -- no argument there. But the question I'm posing to the group is, given the listed steps that he used to create his stills, *why* did that procedure produce uncharacteristically bad images? Specifically, what was he doing wrong? Let's assume that all he did were only the four steps outlined above. What exactly are the mistakes here? No mention is made of pulldown removal or deinterlace technique. Are those the only considerations?

Can we pin down a precise technical reason for this discrepancy though? Or is it just a massive conspiracy perpetrated by HV20 owners and assorted Canonistas worldwide (I asked, tongue firmly in cheek).

Oh, I completely agree with you wholeheartedly. There's no single worse way to "compare" camcorders than by examining still frame .jpg images, it's an incredibly pointless process. I found the fracas about uploading full-size frame grabs to AVS somewhat amusing, as I could easily accommodate those full-size images here, with no uploading restrictions on size or weight. In retrospect I guess I could be talked into doing it, but why bother.

Which leads me to another point that I'm about to bring up in a following post. His public claim that "a video frame grab doesn't have EXIF header data" is false, and I'll explain how and why that is.

Well, not only the HV20 but also the Sony V1U and frankly every other significant HD or SD camcorder on the market -- this is after all one of the largest digital video sites on the web (and the single largest HDV site, bar none). The fact that there is so much good, representative material here is the primary reason why I usually don't appreciate the posting of links to other sites. In this case though, the matter at hand concerns material belonging to Wes and others, taken without their permission and redistributed without proper credit, which most definitely is *not*representative.

It goes back to my question: based on the capture steps listed above, *why* are those frames not representative? What are the technical reasons for the poor quality of those particular images compared to their original versions?

That's a good question. It brings to mind how Mac G3's used to render DV on the fly, it was all pixelated so that it would play smoothly but if you turned on the full quality it was crystal clear. Perhaps there's some low resolution rendering going on in Vegas.

I also agree that judging video from frame grabs just doens't work. Many times you look at a frame grab and it looks blah, but seeing the full motion video can look fantastic.

I have actually been using my video camera to aquire stills, not from the photo mode but actually from video streams because the picture quality is so outstanding. It's like having a 1440x1080 digital still camera that can shoot 24fps continuously which isn't too bad, though the continous shooting still mode on the HV20 is decent also.

Vegas vs other framegrabs
 

Chris, intriguing question. Personally I do most of my framegrabs from MPC. On the HV20, I sometimes use the photo button while recording video but not often. I tried the framegrab method on one m2t with MPC first and Vegas second.

Notes about the m2t:
It was shot with PF24 in TV-48 with my custom effect which Custom effect was Color Depth 0, Sharpness -1, Contrast -1, Brightness -1. Native m2t straight from tape = 24p in 60i transport. Fortunately, the first frame appeared to be progressive.

MPC Method:
1) Load m2t, it will autoplay.
2) Stop
3) Pause (it will automatically go to the first key frame)
4) Save image (options are JPEG and BMP)

Produces a 1440x1080 4:3 image.

Vegas Method:
1) Set project to HDV1080-24p preset: 1440x1080, motion blur: Gaussian, Deinterlace: Blend, Field order: None (progressive)
2) Preview window: Best (full) -- odd thing the image is 1440x1080 but preview makes it 1920x1080.
3) Save file (from the preview window-options are JPEG and PNG).

Produces a 1920x1080 16:9 image

What this tells me about Vegas, the quality of the snapshot will depend on the quality setting in the preview window. Obviously, when size changes from native HDV there will be resampling of the image. The preview window is probably (someone please confirm) also affected by the project blur and deinterlace setting.

Pictures (all 4 formats -- 4.43MB)
http://www.sendspace.com/file/cggbpb

As for the stills, he should really just get permission to use some that other people have taken instead of downloading mt2 clips and making his own from those, no matter how he's done them. To be honest, a few of his V1U stills don't look that good either.

That's my biggest complaint. To not get permission and then post such terrible stills when Wes himself has already provided much better stills of his own footage. It all just seems wrong.

Regardless, this will be my last post on the subject. I think we are giving him far too much attention as it is.

Mike

1920 x 1080 images...

http://www.lousyheros.com/pics/dvi/

Meh... HV10... not true progressive, interlaced (at 1/2000 shutter) converted to progressive... some pics are good some are meh...

Looks 10x as good when you're actually watching the footage... the frame-grabs are a really poor representation

I sent this guy an email and he ignored me, you can read his "latest" update... I guess I wasn't very polite ;) Actually I just called him out... offering 20+ hours of raw footage + as many frame grabs as he wants... he just ignored it

As the originator of this thread I request to the moderators that it be closed.
This guy is getting far too much attention.

Mostly due to the fact that I created this thread.
I just wanted Wes to know that his work was being used in an unfair manner.


Thanks,

Mike

And this will by my last post on this too since I do agree with Mike, this guy is getting far too much attention.

Joe, thanks for the full sized frames. They're here for anyone to see as are many others. Of course you may be accused of doctoring these somehow. :) However it's pretty obvious that the quality is there.

The point is, as I and others have stated before, one of the poorest ways to determine the quality of camera A vs camera B is to use frame grabs. They simply don't do any justice to the full motion video. There are sufficient independent reviews on this camera, resolution tests, sufficient native m2t clips and more than enough user testimonials (some by owners of far more expensive HDV cams that have a solid reference point), to underscore the amazing quality of this unit.

This entire discussion began when I responded to this individual's assertion that the HV20's output was 'soft'. Anyone that owns the camera knows how comical that is. I shot some footage on Saturday at an airshow and the quality was simply stunning. Anyone that would have called that footage 'soft', viewed on a 50" Fujitsu plasma, would have needed emergency treatment at their local opthamologist.

Would I use this camera professionally? Of course not, there isn't enough adjustment flexibility and I wouldn't have the guts to show up at a paid job with a camera of this size. However, does its output compare very favorably to far far more expensive HD cameras in many conditions? You betcha. In the end, isn't that what it's all about? ;)

As you wish, Mike. I agree. This thing has pretty much run its course.

regards,

-gb-