I wish I had seen this thread earlier... If you have any further questions, feel free to ask.
I would avoid transcoding to DV. DV is a fairly old standard definition codec. The main problem you'll encounter using it is the chroma subsampling (color space) conversion will lead to substantial quality loss. It is generally best to keep the files in the native codec, and only transcode once at the end (unless you are using a professional NLE that comes with a good "working codec" like ProRes 422). The way color is recorded in video is different than the way it is recorded in still images. Instead of a RGB or CMYK color space, video uses a color space more similar to Lab Color (if you understand color spaces in Photoshop). Basically, all the "resolution" is recorded in a black and white channel (called the luminance channel), and then color is recorded using two separate color channels (called the chroma channels).
However, because the human eye is more sensitive to light vs. dark (luminance) than it is to color (chroma), the codecs typically compress the color information in order to save space. The level of this compression is described using a number like this: 4:2:2. These numbers describe how much pixel data is recorded per-channel, relative to each other. So, 4:4:4 is "perfect" chroma sampling (because for every 4 pixels of luma, there are also 4 pixels recorded for each chroma "color" channel). Very few things are actually recorded in 4:4:4 though, unless you are using a $100k+ camera. So, the way it typically works is that there is less color information than there is "black and white" information.
For a better real-world example, take the 4:2:0 color space. This is what is used on DVDs. Basically the way it works is this: Imagine your video divided up into square blocks of 4 pixels each (2 tall and 2 wide). The first channel is the black-and-white image of what you see (the luma channel). Luminance isn't subsampled, so each pixel can have it's own brightness (recorded as 4 in the "number"). However, Chroma Channel A doesn't work that way. For that, the 2x2 block can only be chopped in half. Each pixel doesn't get it's own color. You can only "color" in blocks of 2 pixels, so that 4 pixel block can only be shaded two different colors, not 4. Even worse, the Chroma Channel B is actually turned off and there is no color data there at all (dividing the total number of possible colors in half). Then, on the next frame, the two color channels reverse, and Chroma B has 2 "blocks" worth of color data, and Chroma A has nothing. This works because the frames are coming so fast that the eye fills in the missing data. Another example is 4:1:1 used by DV. In this, there are still blocks of 4 pixels. However, instead of alternating between Chroma A and Chroma B, DV requires that all four pixels be shaded the same color (but it allows you the full possible color gamut by allowing access to both channels simultaneously). This is actually made even worse in practice because DV uses a line of 4 pixels for its blocks, rather than a 2x2 configuration, but you can ignore that and just know that converting to DV throws a TON of data away (unless you're already in DV to start with).
There are LOTS of possible chroma subsampling patterns. The most common are 4:2:2 (many "pro" codecs), 4:2:0 (MPEG-2 and HDV), and 4:1:1 (DV). The problem is that when you convert from one to another, you end up getting the "worst" of both worlds. So, if you take 4:2:0 footage and convert it to a 4:1:1 codec like DV you actually end up with the equivalent of 4:1:0. That's because when you convert, every other frame in the source 4:2:0 has either 0 Chroma A or 0 Chroma B.
If you are adding any overlay graphics or anything, you will have much better final quality if your sequence (the edited timeline) is in a codec capable of a 4:2:2 color space. Otherwise you will likely see horizontal banding in graphics (especially anything moving) and "jaggies" around the edges of lines, like fonts. This is because of those differences in how color information is recorded. You can't always have a graphic "end" perfectly aligned on those 4 pixel blocks, so there is often "bleeding" around the edges of sharp lines. If you aren't adding any graphics, then the best bet is to keep everything in the original source codec to avoid color shifts and loss of color data. If you must transcode, always transcode to a codec with a LARGER color space (going from 4:2:0 to 4:2:2 is fine because no data is thrown away).
I don't know if any of this makes sense at all, but I thought I would try to explain...
PS: Here is a nice graphic that makes it easier to understand visually:
http://en.wikipedia.org/wiki/Chroma_subsampling#Sampling_systems_and_ratios
Author
Topic: Any suggestions for software to edit mov files? (Read 24429 times)