smokeslikeapoet
19th August 2002, 08:22
I still had some questions about the greyscale option in XviD, such as, does the color information take up bit space in your encode, and if so how much. So I started with a very low bitrate and worked my way up to find what's and acceptable bitrate. One of the problems is that Cinescope movies (2.35) AR movies are always (i think color) so when you're encoding a black and white movie it takes a higher bitrate to get a desirable quality since your working with higher horizontal resolutions. Test note, I don't test a movie by the number of I/P frames, or the quantizer, I just look at the final product and know what looks good and what doesn't look good.
Here's the setup. Using Pi <http://us.imdb.com/Title?0138704> chapter 14 encoded at 640x384 resolution, XviD 1 pass CBR bitrates 400-600 kbps at 50 kbps increments.
400 kbps: Blocks looked horrible, but the image was viewable
450 kbps: Just a little better than 400
500 kbps: Great improvement, blocks are still noticable but much smaller.
500 kbps: Blocks are almost gone, but greenish and purplish hues appear where the blocks are gone.
600 kbps: same blocks as 500 kbps greenish and purplish hues are much less noticable.
I ended up encoding the movie in a 1 cd 800 MB rip @ 1250 kbps and i expect it to look flawless when I'm done. I plan on preforming the test again on a movie with a wider aspect ration, but i can't seem to find a 2.35 AR black and white movie, perhaps tomorrow i will try 650-850 kbps on the same clip.
Synopsis: I think greyscale, still in it's early stages, is great for saving bits on b&w movies, but pushing the bitrate too low will still cause weird hues and artifacts. I think the power of this codec lies in the alt curve treatment which is not available in 1-pass short clip tests. Perhaps encoding the whole movie at 600 kbps using alt curve optimizations would have produced much less artifacts and a true black and white picture, but time doesn't allow me to run hours worth of tests.
One last thing:
Movie theater screens are HIGHLY reflective (hence the name "silver screen") producing a glow to black and white cinematography. I know that black and white photo paper has silver in it, I think early movie screens did too. Now its probably Mylar. DVD's have yet to produce a technology that would resemble this old analog trick, making the movie viewing experience much less pleasurable. Find a theater locally that plays old classic black and white films and then tell me i'm right. Is there in existance a vdub filter that will produce a glow affect that could reproduce this glow effect in my encodes?
I'm willing to run some more tests if anyone has ideas about how to improve my methodology.
Note: There was already a greyscale thread, but it contained some nastyness that I didn't want revisited. If this looks familiar to you it's because I posted much of it on www.everwicked.com, but this forum is much more active.
Here's the setup. Using Pi <http://us.imdb.com/Title?0138704> chapter 14 encoded at 640x384 resolution, XviD 1 pass CBR bitrates 400-600 kbps at 50 kbps increments.
400 kbps: Blocks looked horrible, but the image was viewable
450 kbps: Just a little better than 400
500 kbps: Great improvement, blocks are still noticable but much smaller.
500 kbps: Blocks are almost gone, but greenish and purplish hues appear where the blocks are gone.
600 kbps: same blocks as 500 kbps greenish and purplish hues are much less noticable.
I ended up encoding the movie in a 1 cd 800 MB rip @ 1250 kbps and i expect it to look flawless when I'm done. I plan on preforming the test again on a movie with a wider aspect ration, but i can't seem to find a 2.35 AR black and white movie, perhaps tomorrow i will try 650-850 kbps on the same clip.
Synopsis: I think greyscale, still in it's early stages, is great for saving bits on b&w movies, but pushing the bitrate too low will still cause weird hues and artifacts. I think the power of this codec lies in the alt curve treatment which is not available in 1-pass short clip tests. Perhaps encoding the whole movie at 600 kbps using alt curve optimizations would have produced much less artifacts and a true black and white picture, but time doesn't allow me to run hours worth of tests.
One last thing:
Movie theater screens are HIGHLY reflective (hence the name "silver screen") producing a glow to black and white cinematography. I know that black and white photo paper has silver in it, I think early movie screens did too. Now its probably Mylar. DVD's have yet to produce a technology that would resemble this old analog trick, making the movie viewing experience much less pleasurable. Find a theater locally that plays old classic black and white films and then tell me i'm right. Is there in existance a vdub filter that will produce a glow affect that could reproduce this glow effect in my encodes?
I'm willing to run some more tests if anyone has ideas about how to improve my methodology.
Note: There was already a greyscale thread, but it contained some nastyness that I didn't want revisited. If this looks familiar to you it's because I posted much of it on www.everwicked.com, but this forum is much more active.