DV is often 4:3 but stored with data aspect ratio of 720x480. 720x480 is *NOT* a ratio of 4:3. DV has non-square pixels and expects to be displayed at 640x480 which *IS* 4:3 ratio. I am unclear whether this is due to legacy technology, or whether there is actually extra horizontal data in DV that goes to waste unless output is displayed at > 640 columns.
DV is interlaced. Interlacing is actually quite common - it was a necessary tactic in order to maximize motion and resolution in early video tech. However handling interlacing is of utmost importance. FCP displays DV footage at correct 4:3 ratio but will export a file that quicktime plays back at 720 causing everything to be short and fat. When converting you'll often have two options - whether the input is interlaced (should be de-interlaced) and whether the output should be interlaced.
http://www.dvcreators.net/interlacing/ is a good introduction
NOTE: besides the various methods of interlacing, there are two fundamentally distinct cases - a case like analogue cameras where alternate fields were recorded at different times (60 samples become 60 fields in 30 frames) and cases where the source was progressive and so both fields in a frame are from the same moment in time - such as a 24p film source being interlaced for DVD. It the case of a film, it is possible to deinterlace and get back the original complete frames. In the case of 60i video, you can deinterlace to 60 frames at half height which keeps the full time resolution. You can then stretch the height by 2 to get 60p - although you can't recover the other half of the lines, you aren't really throwing away data. Result may be a little soft due to scaling, but that's unavoidable. The most common method seems to be throwing away odds or evens, which will convert 60i to 30p - but losing time resolution, as well as the same picture quality as above (half height stretched to twice as tall). In the case of minimal motion or fixed camera, motion adaptive interlacing can identify parts of the screen which are still and combine both frames in those areas, to achieve full vertical resolution. Moving areas can be stretched. This method should work to 60 or 30 frame rate output.
DVD content should indicate the format well enough for player to handle both cases.
Field dominance applies only to interlaced footage. DV "field dominance" is "lower". Most everything else is "upper". If you are working with/to/from progressive video you might have to choose "none". Beware that Compressor presets may have field dominance set to upper, if you are compressing from DV you might need to change it.
CHATTING WITH FRIENDS
eyescratch: FCP will "render" by copying untouched frames. if you don't motion or color correct or timestretch, simple whole-frame editing will be lossless up to point of export. when you deinterlace you throw away half the data. only do titles once you are in progressive!
wait - does that mean every time i render a sequence with motion / filters and render it, I'm taking another hit quality-wise? will final export render work through all the embedded sequences and filters to render each frame fresh? can i control this? do I need to set seq compression to "uncompressed" to avoid it? if i clear all render files before export does that ensure no intermediate loss? if i have interlaced footage in an interlaced seq but clip has motion, then will rendering de- and re- interlace or does it process each field? what is the domain for video filters.
seej: one problem at a time. de-interlace as its own step, don't rely on software to automatically handle it cuz then you don't have control - double true if it is also handling resolution, aspect ratio, etc. this implies export from FCP uncompressed/png interlaced, then deinterlace to an uncompressed, then attack aspect ratio etc.
No comments:
Post a Comment