HV20 and Blocky Images: What's going on?

[Kholi]

Okay I've been wanting to ask this for a while. The images that I'm seeing from you guys, when in a 720 or 1080 sized frame, are pretty blocky. There are a few stills from Forsman in his thread that show the blockiness off quite well.

I was beginning to think this was a drawback of a 1000.00 dollar camera... then I saw Norbert's raw M2T and that blockiness is nowhere to be found.

What are you guys doing in conversion that does this?

http://www.dvxuser.com/V6/showpost.php?p=923308&postcount=17

It's in a lot of the HV20 footage I see if the frame size is 720 or above.

Norbert's (wobbly footage) is the first time I've seen a clean, blockless image.

[rawfa]

Kholi, have it in mind Norbert's footage is PAL (no downconversion needed).

[Kholi]

What about the footage on the Canon site? It exhibits noise, but not the blockiness I'm seeing in the user footage. Also, Norbert's footage samples are M2T's. Is something going wrong in the 24P Pulldown process that people are using?

[rawfa]

That IS curious and you do have a valid point going on here.
Another interesting thing that I notice on Norbert's footage was that it seems to display more detail and it's color reproduction looks more like the A1's than other HV20's I've seen...

[Sacksnack]

Maybe because Norbert posted the original m2t's, and other have been posting WMV and Quicktime? I'm not sure, but his footage, despite the wobble, looked the same as some stuff I shot yesterday. I think if I did a side by side I would have a difficult time telling the difference between my A1 and my HV20, in outdoor shots. In less than ideal situations the A1 wins hands down. I can't stress enough how impressed I am with such a small camera.

Edit- I just read what Kholi wrote about the files being m2t.

-Kevin

[Kholi]

Yeah, that's why I'm asking how you guys are converting the footage. Is this what happens when you do? Or is the process that people are using just all wrong?

The m2t from Norbert's cam is really nice, and made me go "wtf?".

[Norbert]

I don't think there is any magic going on in my footage. I guess it's just the fact that I can shoot 25p and capture it "as is". God I love living in a PAL country. ;)

It was a pretty sunny day today so I had to underexpose and that means that colors become more saturated. In other words, it was an ideal shooting situation because there was more than enough light.
I'm sure footage from my camera would look like footage from any other HV20 in poor light. In good light however, my HV20 blows away all the other HV20's because my HV20 is some kind of horribly mutated monster. Hence the highly unstable footage. ;)

[Kholi]

I honestly think it's the first thing you said. I think the de-interlacing process is somehow going awry.

[David Jimerson]

What "deinterlace" process?

[kyle.presley]

All the footage I have shot looks great. One thing I have noticed, is that when I go to color grade the images, I get some blockies, but not much. It's pretty clean.

[Zak Forsman]

What "deinterlace" process?he's actually referring to the reverse telecine process.

[Kholi]

Sorry, the pulldown, inverse telecine, reverse backflip, juxtaposition proper kick. Either way, it's blocky.

[icicle22]

Some things to consider. I don't know if these all hold true but I think they might.

The HV20 takes a 24P stream and wraps it in a good old fashined 3:2 pulldown. This is not 24PA or 2:3:3:2 either. So, some frames are split across fields and may not be represented as a full frame anywhere. So when converting back to 24P stream from 60i there may be a loss in quality. This was the entire reason Panny had the 24PA feature. To allow the 24P stream to be represented with no individual frames having to be reconstructed using fields.

Also, the HDV codec is compressing this using 60 discreet fields instead of 24 frames like on the H1 and the A1. So the initial codec is far less efficent and will introduce more blockies as you called them. So 3.6MB persecond are split into 60 segments instead of 24 thus leaving less bits for each individual frame.

I typically edit my H1 footage using Premiere and Aspect HD. I render the final HD movie out to HD mpg files, usually around 15megabit stream. I play these acroos the network on my 45" LCD 1920x1080 screen and they look amazing and crystal clear. It is hard to differentiate them from the raw footage as far as clarity goes.

The same process with the HV20 is not as good. I can see some "blockies" here and there and sometimes can perceive a type of vertical edging. In other words certain diagonal lines show a minor stair stepping like pattern simlar to color sampling of DV. This is only prominent in less lit scenes or indoor shots. Outside and well lit scenes do not exhibit this so much.

Anyone else have any insights? Overall, technical points aside, the HV20 just doesn't hold up as well for multiple passes and editing as H1 and A1 do. IMHO.

[Barry_Green]

I think you're spot-on in your assessment. The only other thing I'd say is that some of the "stepping" you're seeing is due to interlaced 4:2:0 chroma. The XHA1 24F uses progressive 4:2:0 chroma sampling, which gives it cleaner footage; in 60i it of course uses the interlaced 4:2:0 chroma. Because the HV20 is always recording a 60i signal, it always uses 4:2:0 interlaced chroma. Interlaced 4:2:0 chroma gets spread across fields and can cause small blocks in the footage that are usually visible only on a progressive monitor or still frame; shouldn't be visible on an interlaced TV. However, there aren't really any interlaced HDTVs anymore, so the benefit of this is dubious.

[Kholi]

So if you applied the pulldown to Norbert's very wobbly footage, we'd still be getting jaggies/blockies or is it because of his PAL cam that it's not going to happen?

The sadness. Cause the blockiness shows up well in 720p 24p footage. At least to me. I know I know-- what do you expect from a 1000k 24p camera.

[Barry_Green]

You wouldn't do pulldown removal to Norbert's footage, because there's no pulldown in it. He's in PAL territory, so it's 25p (recorded as 50i), no 2:3 pulldown necessary.

[Kholi]

You wouldn't do pulldown removal to Norbert's footage, because there's no pulldown in it. He's in PAL territory, so it's 25p (recorded as 50i), no 2:3 pulldown necessary.

Got it, thanks Barry. Sadness it is. Because it looks unrealistically nice at 25p, and the stripped 24p is rather blocky.

Can anyone provide an M2T from a states cam? I'd like to see a raw file to compare to the stripped.

[Barry_Green]

I don't know how people are stripping out the pulldown; I would blame that process before I'd blame the raw footage.

I'll post some raw 24P .m2t here at some point, which will still have the pulldown in it.

[Noel Evans]

I havent had the HV20 long, and I can tell you the image is not as clean as the A1...... in fact it looks more like the footage I got from the Sony V1.

[kyle.presley]

Maybe a dumb question, but could the stairstepping patterns not be caused by the demosaic filter?

[Barry_Green]

Here's a raw 24P .m2t straight off the camera. This includes the interlaced pulldown (obviously). This was cine mode, 24P, 1/48 shutter.

http://www.fiftv.com/Alligator.m2tX

Download, save-as, and remove the trailing X. I don't know how long this'll be up there, that server's short of space already so this might be a temporary thing. But just so you can check and see if you see blockies in this footage; this is again raw from the camera, no recompression of any type.

[Zak Forsman]

frame grab exported with MPEG Streamclip directly from Barry's M2T file of a progressive frame.

http://www.sabipictures.com/heartofnow/gfx/Alligator.tiff

same frame, but Barry's file has been converted to Apple Intermediate Codec, reverse telecined in Cinema Tools to 24p, and the frame grab exported in MPEG Streamclip.

http://www.sabipictures.com/heartofnow/gfx/Alligator24p.tiff

[Kholi]

Thanks Barry. I am downloading it now. Edit here: I don't see any signs of blockyness in the raw M2T.

Barry since you're on a PC, how are you doing the Pulldown? In what program?

Zak I do see a minor "Blockyness" in the second still. It's not as apparent as some of the others, but it exists. Is it a sure thing that it is not due to the method of pulldown?

What are the existing methods in different NLE/Programs? Since Hydra is like months away I'm back to thinking about picking up an HV20 for an upcoming project.

[Barry_Green]

I'm not. I haven't tried to do anything serious with the HV20, so I haven't bothered to keep up on ways to strip out the pulldown. I thought for sure Vegas would have an update that would handle it, but I guess not... I know After Effects would strip it out, if you imported the file to a composition and told it to remove pulldown and then exported as uncompressed or something, but that'd make editing a nightmare.

So I'm just waiting for Vegas to come along with the proper update...

[Emanuel]

Only a 24p pulldown 60i version problem? Or the same @25p footage?

[Barry_Green]

I don't know, but I took those pics from Zak into PhotoShop and put one over the other so I could toggle on/off, and they look nearly identical. There's ever the slightest change in the image from the AIC recompression, but it's so very minor that I don't think I'd even worry about it at all. It certainly didn't make the footage noticeably more blocky or anything like that; there was barely any difference.

[icicle22]

My early tests with the HV20 I am using Cineform Aspect HD to convert to true progressive .avi. The software is setup to remove pulldown as it captures. I have no official idea if this is the best way but usually Cineform is good about this type of stuff. David Newman suggested this himself.

[Emanuel]

Thanks Barry & Zak (and Kholi too for your appreciation). It seems there's no hassles, right icicle22?

Taken your previous input, any potential HV20 user would stay in trouble:

Some things to consider. I don't know if these all hold true but I think they might.

The HV20 takes a 24P stream and wraps it in a good old fashined 3:2 pulldown. This is not 24PA or 2:3:3:2 either. So, some frames are split across fields and may not be represented as a full frame anywhere. So when converting back to 24P stream from 60i there may be a loss in quality. This was the entire reason Panny had the 24PA feature. To allow the 24P stream to be represented with no individual frames having to be reconstructed using fields.

Also, the HDV codec is compressing this using 60 discreet fields instead of 24 frames like on the H1 and the A1. So the initial codec is far less efficent and will introduce more blockies as you called them. So 3.6MB persecond are split into 60 segments instead of 24 thus leaving less bits for each individual frame.

I typically edit my H1 footage using Premiere and Aspect HD. I render the final HD movie out to HD mpg files, usually around 15megabit stream. I play these acroos the network on my 45" LCD 1920x1080 screen and they look amazing and crystal clear. It is hard to differentiate them from the raw footage as far as clarity goes.

The same process with the HV20 is not as good. I can see some "blockies" here and there and sometimes can perceive a type of vertical edging. In other words certain diagonal lines show a minor stair stepping like pattern simlar to color sampling of DV. This is only prominent in less lit scenes or indoor shots. Outside and well lit scenes do not exhibit this so much.

Anyone else have any insights? Overall, technical points aside, the HV20 just doesn't hold up as well for multiple passes and editing as H1 and A1 do. IMHO.

Although with Cineform I bet I wouldn't... Any thoughts?