4.5k & 2k footage in 1080p Timeline

[Ryan Lightbourn]

Hey guys,

I'm sure this has been asked many times before, but I can't find a solid answer anywhere:

I was wondering what the best way to downscale 4.5k & 2k 2:1 footage to insert into a 1080p timeline. Final output will have 2.35 aspect ratio (1080p with 2.35 matte added to any non-2.35 clips).

Do I just, for example, set 4.5k footage to 43% scale & I'm good to go? Or is there a better way to preserve sharpness (especially with the 2k footage)?

Thanks!

 

[Chris McKechnie]

If your'e editing in premier pro, always make sure you don't have "scale to frame size" selected (right click on a clip to quickly check). You can set it as the default under your settings. By manually scaling the clip down, like you've mentioned, it will be sharper than the alternative. I'm guessing it would be the same in FCP.

Oh, and in PP, you can just add a crop/matte adjustment layer to your entire sequence.

 

[Ryan Lightbourn]

Thanks Chris!

 

[Brad Allen]

If your'e editing in premier pro, always make sure you don't have "scale to frame size" selected (right click on a clip to quickly check). You can set it as the default under your settings. By manually scaling the clip down, like you've mentioned, it will be sharper than the alternative. I'm guessing it would be the same in FCP.

I have often advocated using scale to frame size and so wanted to do some scientific testing to make sure I wasn't incorrect or misleading people on this front.

In my testing, scaling to frame size and scaling the clip manually (in the effects panel) produce identical results when using Max Render Quality on export. This is because Premiere does a full debayer and downscale when this option is checked (both for automatically scaled and manually scaled clips). To confirm this, I went a head and created a 2048x1080 sequence and dropped 4096x2160 footage into it. I duplicated the clip. On the first I used "Scale to Frame size" and on the second I scaled the footage to 50% manually. I then exported out PNG stills of the first frame of both clips with Max Render Quality checked and compared the two with a difference Blend mode in Photoshop. There was no difference at all - images were identical. Links to both of those PNG's here. http://unravel.com.au/files/2048x1080_ScaletoFrameSize.png, http://unravel.com.au/files/2048x1080_Scaled_Manually.png

Now, where I expected there to be a difference, and where there was one, is when "Max Render Quality" is left unchecked. In this case the clip that is scaled to frame size automatically appears to have a half debayered process whereas the clip that has been scaled manually continues to have a full debayer process. This is expected behaviour. What it results in is a slightly sharper more aliased looking picture for the clip with the half debayer (in addition to more defined noise) and a slightly softer looking image for the clip with the full debayer. This is also expected behaviour but only once you understand that sharpness does not = resolution. Red have a fantastic write up on this here. The reason the half debayered image look sharper is essentially because it is pixel skipping. Where 4 pixels would be downscaled into one average result, it is simply using one. This is similar to the way a 5D MkII does it's downscaling and can give the impression of sharpness and is also the reason that moire' artifacts and problems can occur in the footage. The problems will be much less pronounced in our case here since the beginning resolution is so much higher than a DSLR, but it's still the same problem in theory. This is also why noise appears less apparent when doing a dull debayer and scaling down. Because the downscaling algorithm has 4 pixels to use to create one, noise is averaged out whereas with a lower debayer we don't have that same scaling advantage.


Now, based on this discussion you may get the impression that I am advocating to never use "Scale to Frame Size" since at times it can result in a lower resolution debayer. However this would be incorrect. In my opinion, using Scale to Frame size is the best option (where is suits) since it gives you the best of both worlds. That is, it allows you to do a full quality debayer when exporting out at "Max Render Quality" and alternatively allows you to export out at lower debayering resolutions when speed is of the essence by leaving max render quality unchecked. For example, using 15 seconds of the clip mentioned above, I compared the speed of exporting the "Scaled to Frame Size" footage with the manually scaled footage with max render quality unchecked on both. I rendered to a h.264 MP4 with a 2 pass encode at sequence resolution (2048x1080). The "Scaled to Frame Size" footage exported out in 1:06, while the Manually Scaled footage took 6:37. A 6 fold increase in rendering time! This makes sense when taking into consideration all that has been mentioned so far. The clip using scale to frame size (when leaving max render quality unchecked) is doing a half debayer and doesn't need to perform any downscaling. The clip with manually scaled footage on the other hand does a full debayer and then still has to downscale that footage to the sequence resolution.

I tend to find myself working in sequences that are set at the source footage's resolution so often scaling is not required, but in those cases where it is and scaling to frame size is applicable, I would recommend going that route.

 

[Kujtim Ereqi]

Brad, thanks for taking the time to test this. I have not done any testing on this topic, but I've been practicing it all the time.. Glad to see somebody confirming it..

 

[Ryan Lightbourn]

Just saw this Brad, thank you very much for that info!

 

[Mike Harrington]

Brad....
Where have you found that unchecking max render quality does a half debayer?

In the adobe docs and on the blogs....the only thing I have found is that max render quality invokes a different scaling algorithm
http://blogs.adobe.com/premierepro/2010/10/scaling-in-premiere-pro-cs5.html

Furthermore....if rocket accelerated, the debayer is always done full....so this is definitely not the case in that situation....
I have never found any official quote that would suggest that max render quality affects debayer at all...

EDIT.....
There is some talk on adobe fourums that max render could control either debayer detail, or debayer resolution....but they also state that it is different with the rocket enabled

 

[Mike Harrington]

Also....averaging 4 pixels in the debayer is very different from the way a 5d does it....
The 5d does line skipping.....it doesn't average those 4 pixels...it doesn't even look at 3 of them....that is why the moire artifacts come into play....

Just want to clear some of these up, as they lead to a lot of confusion