Topics

banding due to Interlaced film transfers in digital

citypound
 

In editing, I didn’t notice it on Vimeo, only on a big screen in full resolution. But much of the project has this ugly video artifact of banding when there’s quick motion.

I’m working on a project using tens of hours of old footage [1910-780s ] most 16 but lots of 8, s8 - some of it was transferred interlaced HD

Is there a way to deal with this other than rescanning?

Thanks Much. Rich Allen Writer Director Rhode Island USA

Adam Wilt
 

> But much of the project has this ugly video artifact of banding when there’s quick motion.

> I’m working on a project using tens of hours of old footage [1910-780s ] most 16 but lots 
> of 8, s8 - some of it was transferred interlaced HD

Many NLEs have deinterlacing options, typically tossing away the even scanlines (and replicating the odd scanlines in their place), or the odd scanlines, or adaptively averaging / tossing where there’s motion tearing (that’s bets, when it works, because you lose less vertical resolution on still subjects), or simply averaging even/odd scanline pairs. I’d check on what your NLE offers, first.

As you found, you need to view the footage at 100% or greater magnification to see what’s really going on. Some NLEs also have settings to display only one field or both fields (possibly called a high-quality settings), so it’s worth making sure that the viewer is set to “both fields” / “full frame” / “high-quality” / etc. 

Adam Wilt
technical services: consulting / coding / camerawork
Vancouver WA USA (no, not that Vancouver, the other one)

James Barber
 

The first thing to do is double-check if the footage is proper interlaced (50i or 59.94i), or an interlaced video that's been baked into a progressive format (24p, 25p, 29.97p). If it's interlaced, there are tools in various NLEs, After Effects and other VFX software, and transcoders to get it de-interlaced via different methods. If the interlace jaggies are baked into progressive footage, you're in for a much tougher time and would probably need to try and fix it shot by shot in after effects or something.

James I. Barber
Director/DoP/Editor
London



On 15 March 2018 at 05:36, Adam Wilt <adam@...> wrote:

> But much of the project has this ugly video artifact of banding when there’s quick motion.

> I’m working on a project using tens of hours of old footage [1910-780s ] most 16 but lots 
> of 8, s8 - some of it was transferred interlaced HD

Many NLEs have deinterlacing options, typically tossing away the even scanlines (and replicating the odd scanlines in their place), or the odd scanlines, or adaptively averaging / tossing where there’s motion tearing (that’s bets, when it works, because you lose less vertical resolution on still subjects), or simply averaging even/odd scanline pairs. I’d check on what your NLE offers, first.

As you found, you need to view the footage at 100% or greater magnification to see what’s really going on. Some NLEs also have settings to display only one field or both fields (possibly called a high-quality settings), so it’s worth making sure that the viewer is set to “both fields” / “full frame” / “high-quality” / etc. 

Adam Wilt
technical services: consulting / coding / camerawork
Vancouver WA USA (no, not that Vancouver, the other one)

citypound
 

Adam, Thanks so much. We tried inputing the interlaced and outputting progressive in time warp in Avid but could not get it to work.
Apparently it was scanned in 29.97 and we’re working in 23.98 is a problem. However we may now rethink resizing all the 1.33 original 
footage to 1.5 wide. as it seems the size up causes much of the banding. Yet it may not be possible to rescan some of this so maybe should 
screen all the old 8 and 16mm mm at 1:33 as well as apply ? whatever we can to get rid of the banding.  We'd been mixing full HD 1:78
digital with 1:66 and 1:33 film transfers and had resized all old film to 1.5 as the jump 1:33 to 1:78 looks too large. 1.5 looks nice. But if we go back 
to original 1:33 film frame size it seems to correct much of the banding. Yet we will still pursue a solution to the banding on what cant be rescanned.

W have original 16.5 mm nitrate film stock converted to 16mm that has the banding and we have replaced with w 2k progressive scans of the 16mm,
But the 8mm is harder to corral, we'd need to rescan it all. Where we cant, we will now down size to 1:33 and apply de-lacing thru a NLE fix. 

Again Thanks -

Rich Allen filmmaker Rhode Island



On Mar 15, 2018, at 1:36 AM, Adam Wilt <adam@...> wrote:

> But much of the project has this ugly video artifact of banding when there’s quick motion.

> I’m working on a project using tens of hours of old footage [1910-780s ] most 16 but lots 
> of 8, s8 - some of it was transferred interlaced HD

Many NLEs have deinterlacing options, typically tossing away the even scanlines (and replicating the odd scanlines in their place), or the odd scanlines, or adaptively averaging / tossing where there’s motion tearing (that’s bets, when it works, because you lose less vertical resolution on still subjects), or simply averaging even/odd scanline pairs. I’d check on what your NLE offers, first.

As you found, you need to view the footage at 100% or greater magnification to see what’s really going on. Some NLEs also have settings to display only one field or both fields (possibly called a high-quality settings), so it’s worth making sure that the viewer is set to “both fields” / “full frame” / “high-quality” / etc. 

Adam Wilt
technical services: consulting / coding / camerawork
Vancouver WA USA (no, not that Vancouver, the other one)


Lars Beyer
 

Yes, de-interlace odd or even fields or field merge.
I have sometimes solved the issue by copying the layer and placing it “on top of itself” in the NLE timeline. Then deinterlace the lower layer with odd field, deinterlace the top layer with even field and finally lowering the opacity/ transparency of the top layer to 50 % merging the two layers. You will maintain the vertical resolution this way as compared to tossing one field away, especially when dealing with steady shots with less movement. The effective exposure time will be twice the original exposure time when merging two fields from an interlaced original.
I have found this method giving better results compared to using preset “filters” across various NLE’s.
Needless to say render settings should be the best available as well as the codec/ file format used for render and output.
I have tried it with FCP7, FCPX & Premiere during “online”.

Best wishes
Lars Beyer
DP/ Colorist
Copenhagen

Bob Kertesz
 

I have sometimes solved the issue by copying the layer and placing it
“on top of itself” in the NLE timeline. Then deinterlace the lower
layer with odd field, deinterlace the top layer with even field and
finally lowering the opacity/ transparency of the top layer to 50 %
merging the two layers.
I've used this method as well, and it works better than you'd think. If
the original interlaced footage has been messed with before you got it
(frame converted, or somehow made into quasi-progressive another way),
it may not be as successful. But it's easy to try it on a small piece,
see if things improve.

-Bob

Bob Kertesz
BlueScreen LLC
Hollywood, California

DIT, Video Controller, and live compositor extraordinaire.

High quality images for more than four decades - whether you've wanted
them or not.©

* * * * * * * * * *