Major Aliasing Problem

[Steve Kalle]

Has anyone seen this level of aliasing from their EX1/3/1r? I have never seen aliasing this bad with either my EX3 or EX1. However, I was recording to my nanoFlash at 50Mb/s L-GOP; so, I don't know if the encoding algorithm used in the nano combined with the higher bitrate can cause this aliasing issue. After seeing what the nanoFlash does when recording a camera with a lot of noise using the 100+Mb/s L-GOP codec, I am curious as to whether the Sony codec can increase aliasing.

Hmmm. Something odd is happening. When I exported a frame from Premiere Pro CS5, the jpeg did not have any aliasing. But, when I rendered within the timeline, there was considerable aliasing when viewed at 100%.

[Les Wilson]

Did you mean to attach a screen shot or something?

[Gints Klimanis]

Steve, are you able to host some clips on a Web server? If you did a parallel SxS and Nanoflash, recording, it would be good to see.

[Steve Kalle]

Originally Posted by Gints Klimanis

Steve, are you able to host some clips on a Web server? If you did a parallel SxS and Nanoflash, recording, it would be good to see.

Unfortunately, I was not able to record to both SxS & nanoFlash with my EX3, which was the main camera angle. My client had not received their 32GB SxS card by the time of the event so I had to use my nanoFlash for the event and use the rest of my SxS cards in the other 2 cameras. There were 2 events in two days and each event was projected to last almost 3 hours each; thus, I only had enough SxS cards for 2 cameras to cover that length. However, I had to set my EX1 to 1080 60i SP (25Mb/s) in order to cover the entire event with just 3 16GB cards in that camera (and a 32GB & 2 16GB cards in their EX1r recording at full quality 1080 30p). Fortunately, the nano and my 64GB Transcend CF cards worked perfectly.

One jpeg was exported from PPro's Program Window using its export button, but it does not show any aliasing which is very surprising.

The other image is from the Print Screen function and then cropped in Photoshop which shows the aliasing.

When I play the video either thru PPro or from a rendered file, I can see the aliasing as well.

I will upload a video to vimeo shortly, but I must leave right now to go vote.

[Jack Zhang]

Check your playback and paused resolutions. Are they at Full? If they are, and the problem presists, try turning off GPU acceleration as your next step.

To truly double check, playback your nanoFlash files in VLC to see if you see the same problem in the originals.

This is not aliasing, this is scaling a half-resolution image to double it's size. This is most likely a Premiere issue.

[Steve Kalle]

Originally Posted by Jack Zhang

Check your playback and paused resolutions. Are they at Full? If they are, and the problem presists, try turning off GPU acceleration as your next step.

To truly double check, playback your nanoFlash files in VLC to see if you see the same problem in the originals.

This is not aliasing, this is scaling a half-resolution image to double it's size. This is most likely a Premiere issue.

Here is a section where I see the aliasing. The source footage is 1080 30p 50Mb/s and this clip is 720 30p CBR 6Mb/s H264. You must have the HD enabled in order to see it. Although, due to the high compression, the aliasing is not as noticeable. When I exported it to DVD and watched on my 46" TV, the aliasing was very apparent. And yes, I do know that some aliasing can be caused by the upscaling, but I haven't seen this amount of aliasing in all of my prior DVDs that were created with XDCAM EX footage.

PS - His suit was actually a greyish black, but my EX3's beautiful IR issue rears its ugly head here. However, I was able to fix it and get it to look very close to its original color by using a couple effects in PPro where I targeted the shadows and midtones. This clip was rendered before I CC'd.

Thanks for the comments so far :)

Alias-720p_CBR-6Mb/s on Vimeo

[Eric Olson]

Originally Posted by Steve Kalle

Has anyone seen this level of aliasing from their EX1/3/1r? I have never seen aliasing this bad with either my EX3 or EX1. However, I was recording to my nanoFlash at 50Mb/s L-GOP; so, I don't know if the encoding algorithm used in the nano combined with the higher bitrate can cause this aliasing issue.

Do you see aliasing when playing back directly from the nanoflash through HDMI to a HDTV?

My guess is the 30p source from the nanoflash is being misinterpreted as 60i. As a result the scaling and display algorithms are working with 540 line fields that are consequently aliased.

Try checking your source files with mediainfo to see if they are tagged as progressive or interlaced.

http://mediainfo.sourceforge.net/en

[Dustin Moore]

Yeah, there's no standard for showing 30 fps progressive on a consumer television so you are
guaranteed to get problems from TV's that expect 60i getting 30p and trying
to deinterlace the result.

If you final product is TV, consider shooting one of the formats that TV's understand
like 1080 30i or 720 60p. The 1080 30p is really a web only thing or a special effect.

Also, I don't think the effect you are seeing is what most people would call aliasing.
You have caused a temporal problem with the fields/frames by trying to pack a
novel video standard into a pipeline designed for an existing video standard.

You can get "temporal" aliasing in a situation like this but it would be because you
were shooting interlaced footage that had so much vertical detail that the deinterlacer
in the TV couldn't properly figure out what was detail and what was motion.

The "spatial" type of aliasing is virtually impossible with the EX1/EX3 unless you
use bad resampling in Premiere which has already been discussed.

[Doug Jensen]

Originally Posted by Dustin Moore

If you final product is TV, consider shooting one of the formats that TV's understand like 1080 30i or 720 60p. The 1080 30p is really a web only thing or a special effect.

First of all, there's no such thing as 1080 30i. That would be only 15 fps if it existed -- which it does not.

Second, you are NOT "guaranteed to get problems with 30P". 30P is now the defacto standard for most high-end broadcast television production in the USA.

If you are doing things correctly, 1080/30P that is displayed or broadcast as 1080/60i should look flawless and identical to the original 30P. You're just taking each frame and dividing it into two identical fields. The result looks exactly the same as 30P. So, the reason that Blu-ray and broadcast television don't need a 30P mode is because 30P material shown as 60i is the same thing.

Believe me, 30P is NOT a "web only thing".

And finally, shooting 720 is short sighted, It's not even HD. It's medium definition and soon won't even be around anymore. I don't even know if you can buy a 720P television anymore unless it has very small screen.

1080/30P -- that is today's standard.

[Marcus Durham]

Originally Posted by Doug Jensen

I don't even know if you can buy a 720P television anymore unless it has very small screen.

1080/30P -- that is today's standard.

720p is very much a standard supported by modern televisions, so there is no need to worry. Unless you are talking about 720p native panels which frankly you don't need to display 720p.

[Doug Jensen]

I guess I didn't make myself clear. Obviously 720P can be displayed on a 1080 TV. Do you seriously think I don't know that? Guess what, I also know that SD can also be viewed on modern televisions. But that doesn't mean that 720P or SD look good.

My point is that most TVs being sold today are full 1920x1080 and 720P is clearly on the way out. A new large screen TV with native 720P resolution would be rare if not non-existent, and that was not the case a couple of years ago. 720P was just a stop-gap resolution who's time is quickly fading away. In my opinion, it is short sighted to shoot in 720P today unless you are 100% sure you'll never want to use it in a 1080P production someday down the road. I prefer to future-proof my footage because you never know what you might need.

Take my word for it, 720P will soon (if not already) have the same stigma that SD footage does today.
I hope that clears it up.

[Eric Olson]

Originally Posted by Doug Jensen

First of all, there's no such thing as 1080 30i. That would be only 15 fps if it existed -- which it does not.

30i by definition means 29.97 frames per second where each frame consists of two temporally independent interlaced fields. 60i refers to 59.94 fields per second which is exactly the same thing.

1080 30i makes sense because the 1080 is the size of the frame and 30i is the rate of the frames. If you want the rate of the fields, than it should logically be 540 60i which specifies the size of the fields and rate of the fields. However, in the quest for larger numbers, camera manufacturers often write 1080 60i which mixes the size of the frame with the rate of the fields.

Now back to the aliasing. If you extract every other line from a interlaced frame you get a field, however, if you extract every other line from a progressive frame all you get is a downsampled image that was obtained using a very alias prone downsampling method. Therefore, it is important that 30p be distinguished from 30i during playback and editing.

[Doug Jensen]

This is really getting ridiculous. 30i is not a real frame rate and does not exist. Please show me one legitimate industry source that recognizes 30i as a real frame rate. If it was a real frame rate, 30i would only be 15 fps. Just as 60i is only 30 fps. It takes TWO interlaced fields to create one frame.
Fields and Frames = two different things.

Frame rate - Wikipedia, the free encyclopedia

I know this thread has drifted off track form the original subject, so I won't post again on this topic no matter how much misinformation is being thrown around in the next post.

[Steve Kalle]

Originally Posted by Doug Jensen

This is really getting ridiculous. 30i is not a real frame rate and does not exist. Please show me one legitimate industry source that recognizes 30i as a real frame rate. If it was a real frame rate, 30i would only be 15 fps. Just as 60i is only 30 fps. It takes TWO interlaced fields to create one frame.
Fields and Frames = two different things.

Frame rate - Wikipedia, the free encyclopedia

I know this thread has drifted off track form the original subject, so I won't post again on this topic no matter how much misinformation is being thrown around in the next post.

Hey Doug,

I appreciate the comments so far, but I know of one reputable company who refers to 1080 30i instead of 1080 60i - Aja. Actually, they list it as "1080i 30". KONA LHi - AJA Video Systems

I do believe the 1080i 30 is correct because the '30' refers to the number of frames per second. 1080 60i is incorrect because it is not 60 frames per second, but 30 fps. All of the 'Video Formats' include the frame size and frame rate, not fields per second; thus, 1080i 30 is correct, and so is 1080i 25.

[Dustin Moore]

Originally Posted by Doug Jensen

If you are doing things correctly, 1080/30P that is displayed or broadcast as 1080/60i should look flawless and identical to the original 30P. You're just taking each frame and dividing it into two identical fields. The result looks exactly the same as 30P. So, the reason that Blu-ray and broadcast television don't need a 30P mode is because 30P material shown as 60i is the same thing.

If you feed a 60i (as you correctly would call 30i, sorry) via HDMI into a modern flat panel TV it
is going to go into a deinterlacer which may or may not do what you claim. Some hardware
deinterlacers that I have seen (some of the Bravia ~2006 era at least) definitely do not interprete
30P in 60i and make a nice 30p screen.

Originally Posted by Doug Jensen

1080/30P -- that is today's standard.

Do you have any references to back this up? If you are shooting un-controlled live action
there is a good bit of judder/loss of temporal resolution with 30p. Not quite 24p but
it's not smooth.