-
Resources
-
Review of Cineform NeoHD for HV20 Pulldown Removal
Published by Matthew Jeppsen October 5th, 2007 in Cameras, Formats, FreshDV, News, Reviews, Tutorials UPDATE: A response from Cineform in the comments below.
The following article is a review of Cineform’s NeoHD utility for removing 3:2 pulldown from the Canon HV20, and an opinion on the quality of the resultant images. Article contributed by Eugenia Loli-Queru.
Cineform is pitched as one of the best providers of a good and well-supported intermediate video format. Among their accompanied utilities you will also find HDLink, a capture and converting tool. We tested the tool for its 3:2 pulldown removal ability with the Canon HV20.
The application weighs just a few MBs and it is easily installable, although you might need to manually “register components” after installation (it’s a provided .bat file). It is comprised by the camcorder control widgets on the upper side of the window and three tabs on the lower side. Through the “Capture” tab you can view which clips are captured from the camcorder, from the “Convert” tab you convert existing captured files from your hard drive to another format, while from the “Export” tab you can write back to the camcorder’s tape.
The camcorder control part is much more useful than on other capture applications as it allows you to manually change the exact time that you want the tape forwarded or rewinded. For example, if you know that a particular scene you want captured starts at 00:00:12:04, you can enter that number to the input form and HDLink will force the tape to “go to” that point.
All three tabs abide to the same user preferences. The preference dialog might be a bit confusing at first, but after a quick glance at the included PDF help, it should be clear as day. For our tests we used the Cineform Intermediate codec, our Desktop as the saving location, .avi as the encapsulating format (we decided over .mov because .avi works better with Vegas), “High” encoding quality with “automatic” frame format, “remove 3-2 pulldown” checked, no rate change, “maintain audio pitch” checked, and de-interlace option unchecked (as the pulldown removal will create progressive frames anyway).
When starting capturing from the tape, the conversion to Cineform’s intermediate codec begins immediately. You will need a beefy PC to get it done in real time, one of these new dual core PCs out there. On my XP with 3 GB RAM, 3 Ghz P4 HT (one of the fastest P4s released), the conversion happens nearly real time, but not exactly. Nevertheless, it’s fast enough.
Once capture and conversion has finished, these .avi files are ready for editing. We used Sony Vegas Movie Studio Platinum 8.0b (VMS8P) and Sony Vegas Pro 8. Unfortunately, VMS8P currently has a bug and it won’t display the Cineform files! This is a major drawback for consumer HV20 users who would like to use VMS8P instead, as Vegas Pro is just too overkill for them (VMS8P supports 24p timelines, even if unofficially, you see). Vegas Pro 8 on the other hand was able to display and edit the Cineform files just fine. The Cineform-produced 24p file had non-duplicated frames, or A/V sync issues, and it was truly progressive. The Vegas preview window was able to preview the Cineform clips without stuttering.
Cineform-high vs Lagarith
At this point, we should compare the Cineform way of doing pulldown removal for HDV to the freeware way, which uses a multitude of open source tools and utilities. Here are the pros and cons of each method:
* NeoHD is able to capture and transform from the tape in one-go. Usually, this is done real time. The freeware way takes about 3-5x the time, depending on the speed of your PC.
* Cineform files edit and playback extremely fast on Vegas. Lagarith files produced by the freeware way are stuttering a lot.
* Cineform files are smaller in size than Lagarith files (about half the size).
* On some few PCs, there is an A/V sync issue when using the freeware method. We haven’t seen first hand any such issue though with either the freeware or the Cineform method in our PC. We think these are isolated instances.
* Quality is visibly better (in our opinion) with the Lagarith YUY2 files than with Cineform set in “high” quality (we tried “film” quality too).
* NeoHD costs $600, or $250 for NeoHDV (lower resolution, less color accuracy), while the freeware way, is, well, free.
Cineform-high vs Lagarith
So, which method is better? It depends if capture/editing/playback speed is what you after instead of slightly better quality and a cheaper solution. You obviously can’t have both, at least not with this Cineform version. I believe both methods have their place in the market and users will choose wisely.
Cineform-high still frame
Cineform-film2 still frame
Lagarith YUY2 still frame
I have to give props to Cineform though because it’s the only tool that will do the conversion during capture, saving lots of time from the users, and also it’s one of the first utilities in the market which supported the HV20 from the get go.
Rating: 8/10
FreshDV guest-contributor Eugenia Loli-Queru is a Senior News Editor at www.OSNews.com. You can catch up with her online at eugenia.gnomefiles.org.
17 Responses to “Review of Cineform NeoHD for HV20 Pulldown Removal”
- 1 Pingback on Oct 5th, 2007 at 2:05 am
- 2 Pingback on Oct 5th, 2007 at 2:06 am
Leave a Reply
-
About FreshDV
-
Sponsors
-
Recent Comments
-
Fresh Links
Nice article yet you have your conclusion about which has the higher quality reversed, so the summation should be a free solution vs something that in general works much better. But here is why your quality conclusions are flawed. You did your testing in Vegas which uses Studio RGB, which places black at level 16 and white at 235 — this is a good idea as that matches the full color range of the camera’s source YUV. While the camera does place standard white at 235, the values 235 to 255 also contain valuable highlight details, which are preserved by using studio RGB correctly, giving a wider image latitude for more natural looking images through color correction. Now your images show that the free approach is chopping these highlight values off, and incorrectly setting the 235 white level to 255–that is why the free approach looks more contrasty. While the default contrasty look may seem more appealing to some it is means all the super whites and super blacks are lost (details in shadows are also crushed away.) The under contrasty image contains all the original details.
CineForm worked directly with the Sony Vegas team to provide the studio RGB support, Vegas actually requests the low contrast, full latitude image from our decoder. Other codecs typically incorrectly assume Vegas wants computer graphics RGB, and introduce clipping. So for a true comparison of the two workflows, you need to place a Computer Graphics RGB to Studio RGB conversion filter on the free workflows output, however doing so will not restore that detailed information. Or, if you want the contrasty look, place the Studio RGB to Computer Graphics RGB on the CineForm clip.
Once you normalize the two clips for contrast. You will see that even the low bit-rate CineForm file contains more detail and texture. Now technically the lossless Lagarith codec should not be reducing image detail, yet the entire workflow must include the decompression for the source MPEG2 data within the M2T files. It seems that the free MPEG2 decoder used is doing a poor job, as the Lagarith image is showing clear MPEG macro blocking artifacts (8×8 squares) and well as muddying the image texture.
Quality is high on CineForm’s design goals, so all elements in the conversion chain need to be working at there best. For the best image quality CineForm also allows you to bypass the MPEG compression stages of the Canon HV20 using a Black Magic Intensity card capturing to CineForm live of the HMDI port. That way there are no MPEG artifacts limiting the image quality.
David Newman
CTO, CineForm
David, I did notice that the Cineform version has less contrast. However, this is not where I based my opinion that I prefer Lagarith’s look. If you look closely at the bulb boxes in the zoomed picture, the whole area around the while box is not as “clean” with Cineform as it is with Lagarith. I see more artifacts in the Cineform-high pic than I see with Lagarith. And I don’t think this is Vegas’ problem too.
Regarding dynamic range I don’t disagree with you. Although capturing in Canon’s HV20 CINEMODE, already provides enough of it.
As I pointed out, that “cleaness” is infact lost information from the original MPEG decoding, the quality of CineForm vs Lagarith compression is not a factor. At Filmscan2 you basically have the quality of a mathically reversable transform, so what you are seeing is the difference in MPEG artifacts of the source MPEG decoder. The free workflow is suppressing a lot of detail, which you have confusing with lower noise, but it is more lossy.
Using the same image example that you site, here is more data from your images demonstrating my point in this picture : http://www.miscdata.com/downloads/freshdv-rebuttal.png
If you wanted the CineForm to look like the Lagarith image, a blur filter when be close, yet it achieve that look without without the MPEG ringing. I did try this, and a Gaussian blur of about 3 pixels gets you close (using GIMP.) Basically the CineForm result is more high def, and isn’t hugh definition the whole point of these cameras. :)
David.
Thank you for the comments, David. Appreciate your insight.
-Matt Jeppsen
“Quality is visibly better (in our opinion) with the Lagarith YUY2 files than with Cineform set in “highâ€? quality (we tried “filmâ€? quality too).”
If you like 8×8 blocks from HDV.
This is a slanted Article…
>This is a slanted Article…
I fail to see where the “slanting” is. I personally, much prefer the Lagarith look. At first look, the Lagarith image looks cleaner — and I know a few people who would prefer it this way. Also, this is why in the sentence you are quoting, there in an “opinion” word.
If you don’t know what “slanting” is, then don’t use that word, because it’s a heavy word. A review is meant to be as objective as possible (and it is), and when an opinion is given, it should be clearly stated as such (and it was).
Eugenia,
I know your history a bit. I love free tools and think they are great.
But running your image with just a bit of color correction shows the HDV blocks in spades…
My history?!? You know nothing. Because if you did know my history, you would know what a huge critic of OSS software I am, and given that Lagarith is OSS…
Look, the whole article is positive for Cineform. Except the quality sentence, where I clearly state that this is my opinion, and that someone else might interpret the results differently (as they did). But instead of actually view the article as a whole, you nitpick in one sentence and you make a mountain out of a molehill. And on top of that, you call the WHOLE article “slanted”.
If Cineform doesn’t find that sentence fair, they had this comment section to get the record straight (and they did so, in a RESPECTABLE way, in contrast to the way *you* entered the discussion). I am always trying to be very fair and objective to the reviews I write (otherwise it’s clearly stated to not be so in cases where things are simply not very clear cut), and this article was not any different.
I am done replying to you.
I did what David suggested and ran the Cineform pic over the StudioRGB plugin. Here’s the result:
http://eugenia.gnomefiles.org/images/versus3.png
Indeed, the Cineform pic has a bit more detail, however, it also has this “sharpen artifact” look too which I personally do not entertain. For some people this is not a clear cut decision as to which one is better — except if you are an encoder expert. I personally don’t see any “HDV blocks” anywhere in the Lagarith image, although I do see a hair less detail and a softer image.
To be honest, most HV20 consumers will never see the difference between the two. Heck, some of them export in .m2t after pulldown and as an archival format after editing, and re-encoding a second pass in .m2t is a magnitude of times worse between the quality differences of Lagarith and Cineform.
I know nothing…Just an breeze in the wind.. It’s all you
Eugenia states in her article that “TO HER EYE” she likes the look of the Lagarith more. And from the images posted, the cleaner image does look better to me too, whether it’s apparently more lossy or not. Now whether adding more post processes would degrade the Lagarith image faster, I don’t know. Sounds like it could, if there’s “less information.” But I like the look of the Lagarith, based on all the images, including the one that David Newman posted to prove his own point. It’s just personal. I also loved the look of Inland Empire shot on a PD150 better than many 35mm features I’ve seen recently. “The look” is somewhat subjective. And fact is none of is will probably see the difference in a moving image on our TVs or computers. What I simply LOVE is Cineform’s ease of use, stability and speed. It’s a great product.
Eugenia’s hard work is invaluable. I haven’t found as thorough and clear a tutorial than she offers on her site. And she’s entitled to say which LOOK she likes better. :-)
In arguing the relative merits of which “look” is prefered, I think that everyone is glossing over the most important aspect of David’s first comment, which is the fact that Cineform retains the recorded super-white and super-black rather than clipping. This is a big deal, folks. You’d be amazed at the shots that can be saved, or at least made to look a whole lot better, by recovering seemingly lost highlight detail. I don’t use Vegas, I’m a FCP user, but I’m guessing that there is a histogram (or levels) display that will demonstrate this. This is *not* the same as using CINE mode. Not the same thing at all. In fact, if you’re the type to shoot in CINE, then you definitely want to make sure that you’re not throwing away the ends of your image.
“everyone is glossing over the most important aspect of David’s first comment”
Nope Didnt gloss it over ….It was said…and thanks for saying it again
CF is more than a codec…Does more things for you than you can see
Thanks for the article Eugenia, keep them coming.
Unfortunately subjective comments such as “the look or by eye” of something will always lead to heated debate. Cineform is a great product and one of the things I like about it is the quantitive analysis David does on his blog….give me hard data every time.
That said when you present to a client or audience they don’t give a toss about the data, it all comes down to the look of the image :^)
I had to laugh especially as there is an article on criticism of films on the Freshdv site, might apply to articles as well ;^)
http://www.freshdv.com/2007/10/how-to-handle-criticism-of-your-film.html
“Cineform retains the recorded super-white and super-black rather than clipping”.
Being a lossless codec, Lagarith will preserve exactly what you feed to it,
including the range below 16 and above 235.
The fact that, by default, Vegas requests 16-235 be expanded to 0-255, is not a fault with the codec.
You can of course do the conversion from YUV to RGB yourself, and give Vegas exactly the range you want.
Cheers,
David.