DV Info Net

DV Info Net (https://www.dvinfo.net/forum/)
-   Open DV Discussion (https://www.dvinfo.net/forum/open-dv-discussion/)
-   -   Color artifacts in Black and White? (https://www.dvinfo.net/forum/open-dv-discussion/82139-color-artifacts-black-white.html)

Alan James December 20th, 2006 04:06 AM

Color artifacts in Black and White?
 
Am I correct in thinking that if you turn your camera to Black and White and lay the B&W image onto the tape that there will be no artifacts cause by 4:1:1 color sampling. Here is an example of what I am asking. I didn’t shoot this but here is the original image.

http://i32.photobucket.com/albums/d1...g/blocking.jpg

There are obviously a lot of red color artifacts. I am wondering if when I turn my camera to B&W will the image look like this?

http://i32.photobucket.com/albums/d1...blockingBW.jpg

Or will there be no artifacts? I have done some tests but I can’t fully tell because I haven’t been able to create bad artifacts. And I am thinking that technically if the compression happens when the data is written to the tape then turning off the color info running to the tape will solve the problem. Am I right?

I want to shoot a black and white movie and I’m wondering if it would be best to shoot black and white in the camera or to shoot in color and turn it black and white later, or will artifacts still be there like in that picture?

Kurt August December 20th, 2006 05:46 AM

it doesn't really matter. dv always stores luminance (white and black levels) uncompressed in one channel and chroma (color) in another. so, in your NLE you should be able to extract just the black and white. for instance, in fcp, if you use a color corrector and set saturation to 0, you loose the compression artefacts. i've double checked this resently. dv and reds: not the best marriage.

hope this helps.

John Miller December 21st, 2006 09:39 AM

In theory, you will be better off recording directly as black and white. DV records video at a constant bit rate. By setting the chroma channels to zero, you free up more space to encode the luma channel. This will reduce the compression artifacts. The amount of extra space available can be quite significant - up to 40%.

If you record in color, you must ensure that the software conversion to black and white is performed in the native YCrCb colorspace and not RGB, otherwise you will get artifacts.

William Gardner December 21st, 2006 11:01 AM

Another consideration on this: while you may be able to reduce the color artifacts by recording directly to black and white, you reduce your ability to modify the "color-to-B&W" transformation later.

For example, in your sample images the red shirt and the blue background both end up looking mid-gray in the BW image. If you recorded in B&W, you're pretty much stuck with this unless you do rotoscoping or something like that. If you record in color, you can first crank up the red (or crank down the blue) before the B&W conversion, which will give you better contrast and "color" separation in the final B&W result.

Bill

Kevin Randolph December 21st, 2006 07:03 PM

A bit off topic to the thread, but...
 
If I were going to be shooting a project that was going to be finished as B&W, I would shoot it B&W for a different reason altogether - more accurate placement of the shades of gray. By shooting in B&W, you would be able to see just how the image will look, if filters need to be added in-front of the lens to adjust the grays, if you're getting exactly what you mean to shoot. Also, the use of color filters could be helpful, like using a blue filter to darken the sky and leave basically the rest of the seen untouched.

Just my thoughts...

Alan James December 22nd, 2006 01:17 AM

Thanx for all the info guys. John I know the logic behind what you said. That is I record directly in black and white it will give me more free data space on the tape for the camera to write the black and white data to, but will it actually write more data to the tape? I always thought that because the luma channel is written to tape every frame anyways that it really wouldn’t give me any more info. Also the camera is limited to the codec it has to output in. So in theory you should be able to record more information but in actuality the cameras don’t function that way. Correct?

Glenn Chan December 22nd, 2006 04:04 PM

A did a quick test, and you seem to be able to get slightly less artifacts with B&W in camera. I tested by rendering out a color image to DV (turning the chroma down in one and not the other)- hopefully this would be similar to what a DV camera would do.

http://www.glennchan.info/Proofs/dvi...0DV%20Comp.psd

2- On the other hand, I would prefer shooting in color since there's interesting things you can do in post.

http://www.glennchan.info/fcpugto/sh...end/shot10.htm

Alan James December 22nd, 2006 04:44 PM

Glenn I gotta tell ya that I actually like the right side images better but obviously both ways were shot in color and then turned black and white in post so I understand that that link was just to show a different method to make black and white images, not the advantage of shooting color over black and white. I am actually going for a look similar to the bar scenes Sin City, closest to the one in That Yellow Bastard. Sin City was shot on 4:4:4 HD so RR didnt have to worry about artifacts. I will be applying a green overlay to the final footage to give it a high contrast monocromatic green look. My main concern is that when you turn the video to black and white in post that color artifacts would still be there but if not then there is no reason NOT to shoot in color. I can always set my camera to black and white when I am setting up my lights and then turn it to color for the actual recording. Thanx for all the help guys.

John Miller December 22nd, 2006 05:41 PM

Quote:

Originally Posted by Alan James
Thanx for all the info guys. John I know the logic behind what you said. That is I record directly in black and white it will give me more free data space on the tape for the camera to write the black and white data to, but will it actually write more data to the tape? I always thought that because the luma channel is written to tape every frame anyways that it really wouldn’t give me any more info. Also the camera is limited to the codec it has to output in. So in theory you should be able to record more information but in actuality the cameras don’t function that way. Correct?

I would expect that it should make use of the space made available by not having to encode chroma data. In DV, the luma and chroma share the same space. DV is encoded in three passes. After the first one, the encoder really doesn't care whether data are luma or chroma - all it knows is that it has a bunch of data (usually) too big to fit the available space. Unless a camera implements a first-pass only algorithm (which is unlikely since the results look awful!), you should get the benefit of less compression for the luma by recording only the luma. (Strictly, it has to record a couple of bytes for chroma - a zero DC coefficient and and end-of-block marker).

The best way to test it would be to try shooting complex scenes in both ways and see if you can detect less artifacts with direct B/W recording.

Glenn Chan December 24th, 2006 03:48 AM

Quote:

Glenn I gotta tell ya that I actually like the right side images better
I think that website is a little confusing, in that the first row shows after/before while the rest shows before/after.


All times are GMT -6. The time now is 08:50 PM.

DV Info Net -- Real Names, Real People, Real Info!
1998-2024 The Digital Video Information Network