Category Archives: Misc.

The DV Format

 

HDV uses MiniDV tapes, but it is otherwise a different beast all together.  HDV info to come.

There are some professional flavours of DV (DV-Pro, DVCam, etc.).  Depending on the exact model, the only difference may be the quality of the camera’s optics (lenses).  In other cases, there may be a difference in the size of the video cassette (larger = longer run times), or the speed at which the tape runs (faster = fewer drop outs (errors)).  But, lets stick to MiniDV/Digital8 which both use the exact same codec:  DV  (DV25, if you want to be exact).

The DV codec works in YUV 4:1:1 colour space, which is explained in detail HERE.

The images captured by your camera’s CCD(s) start off as RGB, and then converted to YUV 4:1:1.  Once this is done, a Discrete Cosine Transform is then applied.  Okay, I don’t really know what that means, but you thought I was pretty smart there for a second, didn’t you?  What it does is throw out more info, starting with what you are most likely not going to miss, which tends to be the finer details.

But, of course, colour sub-sampling and loss of fine detail can only be taken so far before it becomes apparent.  DV footage may look nice at first, but a closer look reveals how much “damage” the DV codec does to the original image.

DV compresses data at a ratio of nearly 10 to 1.  Considering that 90% (!) of the image data has been lost, it looks pretty damn good.

Still, this amount of data loss does cause problems for some things such as green/blue screen compositing.  Take a look at this image.  The colour channels have been separated and displayed as greyscale, so you can better see what’s going on.

Note that the red and blue channel have a “halo” around the subject. This is because the red and blue channels are at a lower resolution, due to the 4:1:1 colour sub-sampling. The green channel fairs a bit better, which is why it’s best to use green screen for chroma keying when shooting on DV.

The DV format is Interlaced and has a lower Field Dominance.  For an explanation of what the hell Interlacing and Field Dominance is, check out the video standards page, comming soon..

Colour Space

Colour Space is a method of defining colours using a numerical system.  There’s lots of crazy colour spaces out there, but lets stick with the ones you are most likely to run across.
RGB

RGB stands for Red, Green and Blue, which are the three primary addative colours.

Addative colours are used when working with light, such as in your computer monitor or  TV.  A blank screen is black and colours are added to produce an image.

Different intensity combinations of these three colours can produce all the other colours of the rainbow.  When all three primary colours are at the same intensity, the result is a shade of grey.

Hey!  Wasn’t I told in school that the three primary colours were Red, Green and Yellow?  Those are the three primary SUBTRACTIVE colours.  Also, the exact colours are actually Cyan, Magenta and Yellow.  See CYMK below.

CMYK

CMYK stands for Cyan, Magenta, Yellow and Black.  These are the primary subtractive colours.  I don’t have a clue why the letter K represents black, but my guess is that they didn’t want to use B for black in case you confuse it with B for blue in RGB space.

These are refered to as subtractive colours because they are used for pigments, such as those you find in a printer.  A blank page starts off as white (all the colours of the rainbow) and pigments are applied to subtract the unwanted colours.

CYMK is mostly used in pre-press aplication and you probably won’t come across it too often in the corse of movie making (except when you’re printing DVD lables of your latest flick:)

An interesting thing to note is that RGB and CMY are the opposite of each other.

You can see this on a colour wheel where: Red is across from Cyan, Green is across from Magenta and Blue is across from Yellow. 

This is a handy thing to know when it comes to colour correcting an image.  For example, if your image is too yellow, it can be cancled out by adding blue (the opposite colour).  And it follows that green counteracts magenta, and red counteracts cyan.


HLS


HLS stands for Hue, Luminance and Saturation.  Sometimes the last two letters are swapped (HSL), but it means the exact same thing.

I can’t think of a file format that saves images in HLS space, but you often find it available in most painting programs, because it’s very easy for people to think in terms of HLS.

You can see how it’s simple enough to pick a colour (Hue) and then modify it to the shade (luminance) and saturation you desire.  Easy peasy.

YUV

Okay, this colour space is harder to wrap your brain around, but it is found in most digital video formats (including MiniDV) so it’s worth understanding.
First of all, I don’t know why they use the letters Y, U, V (must be some big brain scientific reason).  It’s (a bit) easier if you think of it as Luma, -Red, -Blue (denoted as L,-R,-B).  So excuse me, all you notational purists out there while I continue to use L,-R,-B to explain how YUV works…

L,-R,-B capitalizes on a weakness of human visual perception in order to reduce the amount of data used  to create an image and still give the PERCEPTION of no data loss.  The trick is to save less colour information than you do luminance (brightness) information.  The human eye is more sensitive to luminance than it is to chrominance so you can get away with throwing some of it out.

This is done by “sub-sampling” the colour information.  This means that the colour information is captured at a lower resolution than the luminance information.  The sub-sampling ratio is often denoted as 4:2:2 or 4:1:1.  This mean for every 4 pixes of luminance info, there are only 2 (or 1) pixels of colour info.  For example:  Suppose you have a picture that is 100 by 100 pixels in size.  With 4:2:2 encoding, the luminance of the image is saved at a resolution of 100×100, but the colour info (both -Red and -Blue) is saved at a resolution of 50×100.  In the case of 4:1:1 encoding, the luminance is still at 100×100, but the colour is sampled at 25×100.

Of course, before the image is actually displayed, the colour info needs to be scaled up to match the resolution of the luma info.

But where is the colour Green in all of this?  Well, it works like this…
The luma info is a grey scale (black & white) image.  White is made up of ALL the colours of the rainbow.  So if you start with white and then subtract red (-R) and subtract blue (-B), what you’re left with is green. Neat, huh?

This also takes advantage of another attribute of human perception.  Our eye’s are more sensitive to green than to other colours.  In L,-R,-B colour space, green is what is left of the luminance (after red and blue are subtracted) and since the luminance is sampled more often than -R and -B, that means by default, green is sampled more than red and blue.

L,-R,-B (YUV) may seem overly complicated, but (hopefully) now you can see how clever it is in the way it’s geared towards the way human perception works.

Some high end video gear can record in 4:4:4 colour space which means that the colour info in not sub-sampled.  The potentially confusing point is that 4:4:4 is usually YUV space, but sometimes it’s RGB space depending of the equipment specs.