Quote:
Originally Posted by nxmehta
The question is then I guess do the original 720 pixels per line encoded at a PAR of 8:9 contain more information than the resized, 640 square pixels per line that result from a resizing operation after that DAC conversion? I guess I could see the argument that you lose something in that conversion. Kinda weird though  |
I'm unsure why you're referring to "resized, 640 square pixels". Neither NTSC nor PAL generally have 640px width when referring to the standards. 640x480 is just a nice, commonly-supported resolution that is close enough to what the proper square-pixel resolution would be for NTSC such that the error wouldn't matter to 99% of viewers. D1 resolution NTSC is 720x480, period (well, that or 720x486, but this doesn't change PAR or DAR, and in most instances those 6 lines can be cropped from the bottom with absolutely no ill effects). D1 resolution PAL is 720x576, period. When connecting digital video sources (or computers in general) to an NTSC CRT set, the number of pixels that can be displayed is a function of how much time each pixel takes up, the quality of the interconnect, and the sharpness of the CRT's shadow mask or aperture grille. For example, the Amiga running the Enhanced Chip Set can output 35ns pixels and with RGB or S-Video on a high-quality studio monitor, it's possible to resolve most of the pixels on a 1280px wide+overscan screenmode. But, I digress.
If you take an NTSC DVD player, connect it to an NTSC TV, and played an NTSC DVD on it, it's going to be outputting a 720px wide image on that analog signal, period. Nowhere is there a resampling to 640px width done, before or after DAC. If anything, the DVD player might crop to 704px width as that's the true 4:3 picture area, the extra 16 pixels are there for headroom in case there is a delay introduced during digitization (even just 70ns delay will shift the image roughly one pixel to the right).
Basically, if I'm understanding your question correctly, the only instance in which resolution is dropped is when outputting 16:9 material on a 4:3 display, and that's vertical detail, not horizontal.
Personally, I always preferred to just keep 720x480 material at 720x480 and slap a 4:3 DAR on the file. Especially in the case of interlaced content, messing with the vertical dimension is just asking for trouble (either causing the field order to flip if you crop an odd number of pixels off the top, or causing distortion of the interlaced field structure if you resize vertically at all). If you insist on keeping 712x480, I'd suggest option 1, since an exact 4:3 DAR doesn't matter so much as the metadata being accurate to the image content.