Setting the Sharpness for getting a film look

[Marco Iannaccone]

Thanx again for your answer! :-)
You're alway very clear, and your posts are beying very useful to me.


<<<-- Originally posted by Barry Goyette : Frame mode on the gl2 is less afflicted with the problems seen on the xl1, gl1 and even xl1s. -->>>

Considering I wouldn't use "stange" shutter speeds (I wanna have a film look, so... I'll mainly use 1/50 s, at least for all "normal" shoots, and maybe 1/25s and 1/100 for getting particular effects on motion clarity, but... almost never), averything should be good with frame mode.... the quality should be at the same level (or almost) than FieldsKit or Magic Bullet deinterlacer, but.. with really less problems... so... it seems the bes choice is actually shooting in frame mode... right?


This won't obiously be the only thing to use for getting film-like look.
I think I'll do some esperiments on lowering the sharpness, but I think 2-3 points down could be a good start point.
I'll obiously take care of exposition, avoiding large overexposed areas, and doing a good white balance, and... will take care to fast camera moves.

I read somewhere, for getting good camera moves with light cameras like XM2, I should move it as it were actually less light... as it would actualy be a bigger one... avoiding extremely fast movements, etc...
Obiously some sort of support must be used...


<<<-- You will need some discipline when shooting with frame mode. Wildly eratic camera movements (especially hand-held-rotational-squirrely ones) don't look good in any situation, but they can really be annoying in frame mode (Film is even worse). Panning at moderate speed across high contrast vertical objects will cause a strobing effect that can be particularly ugly. -->>>

That's what I was in fact considering.
Anyway... this same "problems" would appear even when shooting interlaced and then deinterlacing, arriving to 1/25 p, instead than 1/50 interlaced, but with frame mode there's te advantage I can directly see (through an external monitor, not the LCD or the viewfinder) when a camera movement causes problems.


<<<-- And definitely do a search for Frame mode here on dvinfo.net...this subject has indeed been covered to death over the past few years....and I'm sure there areat least a few bits of wisdom to be gleaned from it. -->>>

Thanx, I'll do that! :-)

[Barry Goyette]

Dave

The loss of resolution in frame mode is indeed 25% in the vertical direction only (I don't know where the "to 30%" comes from), however, if you read canons marketing, they swear it is in fact a 25% increase in resolution. Who's right....well, it depends on your point of view. Each field of frame mode footage does have 25% more resolution than an interlaced field from the same camera. However a frame (two fields) of frame mode, has 25% less vertical resolution than a frame of interlaced footage.

Confused yet...well when I say that frame mode is only nominally less sharp than interlaced mode I'm really talking what you see with your eyes. NTSC video is very low in resolution to start with, and the detail is rendered in alternating scan lines, effectively halving the resolutiion in any moment in time. The 25% decrease (or increase) essentially means 1/2 of a line...well this is difficult to ponder in my mind, and in fact the real world result is negligible, and certainly made up for by the gl2's higher native resolution and sharpness controls. On top of it, we are talking only about vertical resolution...which means that things like a stack of toothpicks might suffer in the end product, but most other random types of detail will suffer not a bit.

On top of this, you have to realize that this is probably the way it should be for display on an NTSC monitor or TV set. This bit of blurring helps to eliminate a tell tale sign of too-much-resolution 480p footage (like that from the DVX100). In some ways, when viewing on a regular tv...my gl2 footage looks more filmlike than stuff from the dvx, because it has none of the line twitter (a sure fire digital artifact if there ever was one). Sure, the DVX has a way of dealing with this...called the thick detail setting...which effectively lowers the vertical resolution exactly 25%.....hmmmm.

Indeed when viewing on progressive output device like a projector, lcd monitor, or computer monitor, having the full 480p available is a boon to those who have it. But the gl2 and other canon cameras were designed specifically as NTSC and PAL devices, and thus I think that the Frame Mode is in fact slightly lower in resolution "by design". Remember that the DVX100 which came along several years later is really a product designed for digital cinema..ie eventual transfer to film, or digital projection thus it benefits from from all the resolution of its 480p capture.

Barry

[Marco Leavitt]

I realize much smarter people than me are always citing this 25 percent vertical resolution loss figure, but it just doesn't make sense. On my television, which has been adjusted to display anamorphic footage in its correct aspect ratio, the diference between digital 16:9 and using an anamorphic lens adapter is striking. Even though you lose some horizontal resolution with the adapter, the footage is much sharper because you retain that 30 percent of the vertical resolution that would otherwise be thrown off. As I said before, when switching to frame mode the difference isn't just slight -- it's nonexistant. I've tried switching back and forth and asking other people if they can tell the difference and they can't. I can accept that the 25 percent figure makes sense on paper, but it just doesn't seem to be true in practice. I'm using a GL1 by the way, which is much less sharp than the GL2, so maybe that has something to do with it.

[Barry Goyette]

I'm not sure I would agree that it is non-existent, but depending on your monitor, that may in fact be the case. The visible loss of sharpness is subtle, but apparent on the gl1. It is significantly less visible on the gl2. But, as you say...the 25% on-paper loss significantly overstates the actual visible loss, and most people couldn't tell the difference unless you showed them what to look for.

Barry

[Marco Leavitt]

In saying the difference was non-existent, I was referring to my television, which is admittedly a piece of junk. I've considered that maybe the resolution is so low on my monitor that a loss of 25 percent vertical resolution wouldn't even be apparent, but then why is the difference in anamorphic footage so noticeable? Digital 16:9 is known to reduce the vertical resolution about 30 percent. If frame mode was really costing you 25 percent of your vertical resolution in any meaningful sense, I'm certain the difference would be visible on my TV. Why isn't the combined use of digital 16:9 and frame mode the disaster that it would surely have to be if Steve Mullen is correct? I'm not trying to say I'm smarter than him, because I know I'm not. In fact, I don't understand his article at all. It makes my head swim.

[Barry Goyette]

There are definitely two different things going on here. In camera stretching (actually cropping and then blowing up) is a whole different issue than the pixel shift line creation algorithm of the frame mode. And on a camera like the gl1 which, with a 270k chip, is barely meeting the dv spec, its understandable why you would easily see a difference between the 2 16:9 methods. This difference is less visible with the gl2, and I have read some discussion as to whether there is much visible difference between the two 16:9 methods.

I may have read steves article awhile ago, but don't really remember the specifics...but I can see where doing in camera 16:9 (not anamorphic), combined with frame mode, on a gl1, or xl1 would give some people the heebee jeebies.

In my experience with the gl2, and dvx100, I have used the crop and enlarge method, played back on 34" Sony Wega HD set, and it looks really good...occasional jaggies, but no real lack of sharpness.

Barry

[Marco Leavitt]

Ah, I had a feeling it would be something like that.

[Marco Iannaccone]

At this point, I have two questions:

1) Above artefacts with moving parts of the image when using deinterlacing algorithms in post-productions, does it happen any kind of "relosution loss", respect to the original, interlaced image?

2) When using frame mode, considering the image compressed to DV formatcomes non-nterlaced, do the artifacs of 4:2:0 subsampling (I'm obiously referring to the PAL version) due to interlaced footage (strange chroma values in certain points, etc..) tend to desappear, as I would be normally recording real progressive footage, right?

[Barry Goyette]

Marco

Deinterlacing programs like fieldsKit use a variety of schemes to handle the misalignment of two interlaced fields (when the subject is in motion). Any of them will either cause a significant loss of resolution, or some sort of artifact.

Question 2 I don't think I understand...and I'm not sure I'd have the answer if I did.

ciao

Barry

[Marco Iannaccone]

<<<-- Originally posted by Barry Goyette : Marco

Deinterlacing programs like fieldsKit use a variety of schemes to handle the misalignment of two interlaced fields (when the subject is in motion). Any of them will either cause a significant loss of resolution, or some sort of artifact. -->>>

So... what whould you advise in general... to use XM2 frame mode instead of later deinterlacing?



<<<-- Question 2 I don't think I understand...and I'm not sure I'd have the answer if I did. -->>>

Well... read this:
http://dv.com/print_me.jhtml?LookupId=/xml/feature/2003/wilt0603

PAL uses 4:2:2, NSTC uses 4:1:1 subsampling, but MiniDV uses 4:2:0 for PAL and 4:1:1 for NTSC.

Considering he better vertical resolution and other factors, 4:2:0 subsampling used for PAL DV would normally be a very good compromise, but... when having to do with interlaced wideo, it can take to strange color results in centain cases, in particular on edges.


That's why I asked:
If I shoot 4:2:0 interlaced and then deinterlace, I can incur in these color artifact problems, while shooting in frame mode should eliminate these probles, just like as I were shooting true progressig.

Do you think I'm right?

[Barry Goyette]

<<So... what whould you advise in general... to use XM2 frame mode instead of later deinterlacing?>>

I think that's what I've been saying all along.

Regarding the 4:2:0 thing...I don't know enough about this issue to say anything intelligent...Realistically I don't think there would be a significant reason to choose one over the other (deinterlacing method) because of the sampling method. But someone with more knowledge about this should jump in if I'm wrong in saying that.

Although now that I think of it, here's a relevant issue...frame mode occures prior to compression, versus a smart deinterlacer, which must recompress the data after processing...this has nothing to do with the sampling method, but my experience with DV has shown me that the compressed color space it has is extremely fragile...ie you don't want to recompress if you don't have to....another reason to just stick with frame mode.

Barry

[Marco Iannaccone]

<<<-- Although now that I think of it, here's a relevant issue...frame mode occures prior to compression, versus a smart deinterlacer, which must recompress the data after processing...this has nothing to do with the sampling method, but my experience with DV has shown me that the compressed color space it has is extremely fragile...ie you don't want to recompress if you don't have to....another reason to just stick with frame mode. -->>>

Well... right.
And... by the way.. YUV subsampling happens at the very first step of Dv compression (that is: RGB to YUV conversion, YUV subsampling, the rest of the DV compression), so... I really think, because the image come progressive, the subsampling part can produce an image without (or with less) aftifacts.