Compression Artifacts! What Compression Artifacts?

E

eliotbaring

Well-known member
The sony HD-FX i ordered for the school i teach film at turned up today and after doing some testing with my main man Ivan, which involved lots of shakey cam and quick zoom in and outs i could not see any so called artifacting.
What exactly does the HDV artifacting look like and under what conditions is it most noticeable?

For the money it's a great little camera but i only went for it because it's true 16:9, i prefer my DVX over it any day.
 
HDV artifacting is overrated, it hasn't proven to be as much of a problem as everyone here claims, plenty of good work has been done in hdv.
 
In fast movement, the image quality is lower. The question is if the eye can see that. It can defintely see that is frozen frames. Take a look at the following examples. All 400% zoom.

The first is from Kaku Ito HDV clip using HC1. Bob deinterlacing to avoid deinterlacing artifacts. Other that this loss, this is the quality you get on tape.

The second is from same clip. Bob deinterlacing to avoid deinterlacing artifacts again.

The third clip is from DVCPROHD png that was recently posted here. Scaled to 1440x1080 for comparison.

HDV can be wonderfull in scenes without motion. The second clip has less motion and is ok. The first clip is on a Kaku Ito bike trick and it loses quite a lot of detail. The bob deinterlacing reduces only vertical resolution so we can't blame it for the HDV problem. Both examples were Bob anyway. The HVX200 is obviously a lot less compressed but in a competely static scene, HDV would be much better! The real world CF25 resolution of HDV on the FX1 after some tests we did was from 1300x500 to 600x400. Hard to believe but that's the truth. When most of the frame changes everything turn to blocks. That does not happen in typical filming though. Only 5-10% changes usually and HDV can handle that. But a fast action scene is practically SD resolution or lower. Anyone that tried to greenscreen something with complex motion that covers the full frame is aware that it is almost impossible to avoid chroma blocks.

 
Last edited:
Here is a Canon shot. Direct from m2t. No deinterlacing. Focus is fine and exactly on the car. 400% zoom. Camera is panning to follow car. In camera quality. HDV again from the "highest resolution" HDV camera according to recent test:) You can see the blocking and also the "Canon noise" which is not really noise but compression artifacts. Any scene, no matter how well lit, has this artifact. You can see it on the dv.com 6 camera shootout.

 
Another Canon XL 1H frame. Notice the blocky quality of the blur from the panning. This is 100% from 1440x1080 frame. Not 400% zoom like the other examples above. Still quite visible. Color only makes it worse.

 
If you want to see "what compression artifacts" HDV can introduce, check out this series of shots. These were taken on an FX1 at the "Sirens of T.I." show on the Las Vegas strip. When the explosions went off, HDV became very, very unhappy.

These are direct imports from the timeline, exactly as they appear on the tape. No zoom-in, no magnification, no post-processing of any type, and no recompression, they're stored as lossless .PNG's straight from the FCP timeline. This is native HDV at its worst. Normally HDV looks better than this, so you can consider this a worst-case scenario. Look at the lego-sized bricks on the back of the ship, look at the smeared no-definition palm trees in the upper right, etc.

http://www.fiftv.com/FX1/Image5.png
http://www.fiftv.com/FX1/Image6.png
http://www.fiftv.com/FX1/Image7.png
http://www.fiftv.com/FX1/Image8.png

Probably the worst of the bunch is Image5. Here is an extraction from frame 5. Note: this is NOT zoomed in or magnified in any way! This is 100% native size, a direct extraction from the frame. This is the kind of macroblocking that can (repeat: CAN) happen in HDV. It normally doesn't, of course, normally HDV looks a lot better than this.

Image5-Extraction.jpg
 
Last edited:
I think you all helped me make up my mind. I was on the fence between the b and the fx1. After seing those stills i think Im going with the b. Also because of the pour audio imputs on the fx1.

Thanks guys

EVS here i come
 
To be fair, the HVX has its share of blocking. But is harder to see it. HDV under heavy motion can lead to huge contrast at the edges of blocks whick makes them stand out.
 
I hope you don't mean the DVX100B NM_film_maker:)

THE DVX100B has got 4pixel color bleed and 180 horizontal chroma pixels and lots of blocking in itself. They just show less since the image is very small. It is certainly inferior to SONY FX1 cf25 downsampled to SD res. By FAR.
 
I figure I'll make it easier on you by showing you what the FX1 looks like normally, vs. what can happen to it when the compression gets overloaded -- the type of macroblocking that can happen. Note: these frames are not taken from the same moment in time; I just grabbed a shot from the FX1 that had a lot of light (blue light, but still) to show what it looks like when it's not macroblocking.

FX1-vs-FX1-Ship.JPG


Again, for clarification -- HDV doesn't normally look this bad. This is an example of what can go wrong when the compression gets overloaded.
 
NM_film_maker said:
I think you all helped me make up my mind. I was on the fence between the b and the fx1. After seing those stills i think Im going with the b. Also because of the pour audio imputs on the fx1.

Thanks guys

EVS here i come
Click to expand...

You can spend 100$ to get two XLR inputs.
 
Beautifull photoshop art , I love it! The blokiness, the over exposed parts, the color rendering, what filters did you use? :)
 
Last edited:
myfriendimage said:
Barry, I think those shots are way bias. The Fx1 in that shot is getting the short end of stick, the backround is way over exposed and not set when the HVX seemed to have taken the time to focus/set iris/ etc...
Click to expand...
Please read what I wrote -- those shots are not meant to be direct comparisons! These shots are not from the same moment in time, and the overexposed background is due to an explosion going off (which is what caused HDV to choke).

I posted the HVX shot so you could see what the detail would have looked like had the HDV compression not turned it into a sea of macroblocks. I said that at the top of that message; you even quoted that part -- so how do I get accused of bias? Does nobody read what I write on these things? I try to spell out exactly what's being compared, exactly what happened and why, and exactly why a certain shot is included. For reiteration, the HVX shot was included to show what the ship's detail WOULD HAVE looked like, had HDV not choked. So I chose the most brightly-lit scene to show the detail.

There is no reason to draw comparisons between exposure or anything else like that. That's not the purpose of showing that shot.
 
xray said:
Beautifull photoshop art , I love it! The blokiness, the over exposed parts, the color rendering, what filters did you use?
Click to expand...
I used a stock Sony FX1 on stock default settings, that's what. You can attribute the "over exposed parts" to the fireball of an explosion going off, and you can attribute the "blokiness" to HDV having a fit when the explosion went off.


Thanks for calling me a liar, that's really appreciated. I can post the original .m2t transport stream if you want. That shot is absolutely unmodified from what the FX1 recorded. And, I posted where the shot was taken so anyone is free to duplicate it to see that I did nothing to cause it.
 
So, the quality dropped for a few frames. The area on the right is entirely black on the HVX so we can't really compare anything. Can we have an FX1 frame without an explosion or an HVX frame with an explosion?
 
You must log in or register to reply here.
Back
Top