When Neo Scene detects 4:2:0 chroma it properly interpolates the source chroma to 4:2:2! Hi what does this mean? if my p2 avc intra 50 source is 4:2:0 can you restore the chroma colors?

nope.....it fills the blank spots with null bits

...sort of like my brain does at 4 am...

So what you're saying is that transcoding from 4:2:0 to 4:2:2 provides no benefit? I presume these null bits do not impact the image and just take up hard drive space. Awesome.

Thay have absolutely, completely and terminally, NO effect, whatsoever on the capture images.
Disclaimer: When you get the data into your NLE, and, you start pushing those bits around(null and otherwise) in transitions and CC and FX, suddenly, you'll find those null bits aren't so null anymore. Now what?
The trick is to get those extra 4:2:2 bits out into a delivery format.
Strictly my observation, when you start with data that has more precision, the roundoff error in the delivery format is less rounded off. Gobbledeegook? perhaps

My take on it also has been that CF extrapolates the data to 4:2:2 from 4:2:0, and that it does make a difference when filtering, tweaking, etc., and in particular if you are starting with HD 4:2:0 and want to end up in DVD 4:1:1, the "4:2:2" DI really helps.
Myth, or reality?- I don't know, but it looks real to me.

Bottom line is that CineForm didn't go for 4:2:2 just because they felt like it. I'm sure they could have made a 4:2:0 version of their codec if they wanted to - it would be faster and file sizes would be smaller. Clearly there are advantages elsewhere in the workflow to make the engineering effort worthwhile.

nope.....it fills the blank spots with null bits

I hope David chimes in here.

I see clear examples of Cineform 4:2:2 filling in (interpolating) missing color samples, especially in the reds and oranges. So how do you classify adding Red pixels to round red objects that appear MUCH smoother than the original HDV or AVCHD files?

I see clear examples of Cineform 4:2:2 filling in (interpolating) missing color samples, especially in the reds and oranges. So how do you classify adding Red pixels to round red objects that appear MUCH smoother than the original HDV or AVCHD files?

I'm not sure what your reasoning process is, here. One cannot create something from nothing. For example, if you're on the edge of an object, interpolating the missing color data can be very incorrect. I would be very surprised if Cineform did this. It would just be wrong. Once again, at the risk of being repetitive, one cannot create something when nothing is there to begin with.

My reasoning process is visual observation...Cineform adds back in the Color information. Double click on the image to see.
Plus on the their website the clearly state the interpolate 4:2:0 up to 4:2:2

Send an email to David Newman. He can explain it much better than I can.

I have many more samples if needed. Take a look at your own Cineform footage and compare with the original.


PS-you can create something that is not there. If I give you two points some distance apart, you can creat a line.

You're still missing the point. Of course you can draw a straight line between two points, it's called linear interpolation. The question is, is it right? In physics, lines are rarely linear. They could be power functions, or an infinite number of other permutations. To assume it is linear is generally a bad guess. And that's all it is...a guess...and not even a good guess.

Well, whatever they do, they do it very well. DVDs make from CFHD intermediates always look better on my HD TV than DVDs made from the M2T (I've had to make a couple that way due to the CFHD vs. CS4 duel... :( ).

As far as the end customer is concerned, it's what shows up on the screen that counts. You can run pixel comparisons on frames, but all that does is give you a number. The final image on the delivery mechanism is the important part, and in my opinion (however invaluable it may be ;) ), Cineform greatly improves the final product...

Thanks,
Matt

nope.....it fills the blank spots with null bits

Bill, I think there is a good chance that this isn't the case.

First off, the bits have to be filled with meaningful data, so they won't be null. At a bare minimum, the additional chroma sampling woud be filled with horizontal and vertical data from adjacent pixels. They're not going to be left empty or null, the image would look like crap if that were done.

But that's being a little picky, b/c what you're really saying is that Cineform doesn't interpolate to fill in the data. But from what I've come to understand, codecs do interpolate quite often. I know that Avid's DV codec does interpolate, while Apple's does not. This makes Avid's good for final encoding but Apple's better for multigenerational use.

Also ten-bit DNxHD dithers and interpolates the 8-bit values. It doesn't just just remap the 256 values to their 10-bit equivalents. If it did that, 3/4th of the possible color values would be left blank. To confirm this, look color samples of 8-bit material transcoded to DNxHD 10-bit. You'll see a lot of values that aren't divisible by 256.

Now I know the argument that the extra space is there for the material to "grow into" while doing color correction and adding effects. But I believe there is a good chance the transcoding fills in the extra chroma resolution with interpolated data.

My reasoning process is visual observation...Cineform adds back in the Color information. Double click on the image to see.
Plus on the their website the clearly state the interpolate 4:2:0 up to 4:2:2

Send an email to David Newman. He can explain it much better than I can.

I have many more samples if needed. Take a look at your own Cineform footage and compare with the original.


PS-you can create something that is not there. If I give you two points some distance apart, you can creat a line.

While I think Cineform does interpolate, I obviously can't confirm it. I do know that Cineform also perfoms a type of de-artifacting routine, which could also account for what you're seeing.

Well, whatever they do, they do it very well. DVDs make from CFHD intermediates always look better on my HD TV than DVDs made from the M2T (I've had to make a couple that way due to the CFHD vs. CS4 duel... :( ).



I NOTICED THE SAME THING!!! I just got my Canon HF200 and am just doing trials of Cineform and messing around with Elements 7.

I found that Cineform does a great job of adding some color, sharpening my image, and just keeping it clean, compared to my using (and even viewing) straight M2T. Mind you, this is just from visual observation, but I'm glad to see I'm not the only one to think so.

Why is this? I instantly became an Cineform Neoscene convert based on the quality of the AVI alone. I'm surprised that M2T looks so bad on my computer.

My dilema now (still testing) is how to edit my video and export it out, for DVD's or youtube/video sharing and keep the same clean and crisp quality as the Cineform conversion.

It seems the moment I try to render the video into anything else, the video compression is very noticable, the edges get soft, lose some color....

My dilema now (still testing) is how to edit my video and export it out, for DVD's or youtube/video sharing and keep the same clean and crisp quality as the Cineform conversion.

It seems the moment I try to render the video into anything else, the video compression is very noticable, the edges get soft, lose some color....

I used to use Procoder 3 & Sorenson Squeeze for final compression, but lately I'm finding that the CS4 version of Adobe Media Encoder is doing a great job of compression for BR, Flash, and DVD.
Another tip for DVD output: If you are in CFHD.avi 60i, it is better to downscale the HD master .avi to a CF SD 60i .avi master and make your m2v transcode from the SD master.
Cineform HD link does a superior job of downscaling interlaced CFHD to SD.
You can really see the difference.

One cannot create something from nothing.

hvx200's do it for a living.

think of intelligent image resizing, spatially interpolating slow motion plugins, etc. There's a ton of pretty well written algorithms out there that do, in my opinion, a darn good job of calculating something from nothing. Why there shouldn't be one for color sampling is beyond me.

Also, color interpolation is a very new and interesting idea to me. While thinking about interpolation in general, I got to thinking about if anybodys written anything to get smoother results out of 4:2:0 footage, and some googling led me to this post. Looks like some software is already doing it.

Also, I found this pretty interesting article - Improved Color Subsampling (http://www.glennchan.info/articles/technical/chromata/chromata.html)

He talks about how programs are already interpolating compressed color, and how he has written his own algorithm that's doing it much better. Look at how much he recovered from the dv compressed image, crazy

No comments?