Log in

View Full Version : Mod16 question and DAR question...


alc0re
31st October 2008, 07:09
Ok...

I have recently purchased an iPod Touch (love it by the way) and I've been teaching myself a few things about converting videos to different formats that the Touch supports.

I've learned a lot but I still have a few questions.

First, my questions are general questions about video and not necessarily specific to my iPod Touch, so I decided this would be the best forum.

Background : My source of video has been backups of my own dvds and captured high def video from my cable DVR (Motorola DHC3416) captured using CAPDVHS. I've been using the following tools : MeGUI, mpeg2repair, DGIndex and VideoReDo.

Question 1 : First question is about resizing as it relates to Mod16. What I want to know is this : What am I loosing by not resizing to a number divisible by 16? At a given bitrate, will video look better if its resized to a Mod16 value? I know some codecs don't support non-Mod16, but x264 seems to. I know some players dont decode/display non-Mod16 video correctly, but the iPod Touch seems to. So what am I loosing if I choose to resize to a non-Mon16 value? Assuming I am loosing anything resizing to a non-Mod16 value, does it get worse the further away from divisible by 16 it gets? What I mean is, if I encode to a number divisible by 4, am I loosing more than by choosing a number divisible by 8? And divisible by 2 worse than divisible by 4? I'm very picky about how my video looks on my iPod, so I really don't want to loose any quality if I don't have to, but sometimes I want to resize to resolutions that are non-Mod16 to try to preserve the original aspect ration. I know I can use a DAR to preserve that aspect ratio, which leads to my next question...

Question 2 : I made 2 encodes of the same short clip of a movie. The first one, I re-sized to a resolution that was fairly close to the original aspect ratio, and DID NOT use a DAR (aka Clever Anamorphic Encoding in MeGUI.) I encoded it again, but this time I DID use a DAR to make the aspect ratio exactly match the original source's aspect ratio. My theory was that since the pixels have to be stretched during playback when using a DAR to get the correct aspect ratio, that the video would look slightly more blurry than one encoded with square pixels (ie my first test encode. ) I captured two stills of the video on the same frame, and to me it did appear that the video using a DAR to obtain the exact correct aspect ratio did look blurry compared to the video that had a slightly-off aspect ratio but encoded with square pixels, since no stretching of the pixels was required (other than the resizing, which was done to both samples.) My question is...is that conclusion/theory correct? Does using a DAR on video make it look blurred compared to square pixels? Like I said, I'm picky about quality, but I also want the correct aspect ratio. Hence my first question. If using a DAR does cause the video to look less sharp/slightly blurry, I would like to encode at non-mod16 so I can get to within 1 pixel of exact original aspect ratio using square pixels.


Thanks

LoRd_MuldeR
31st October 2008, 11:57
Q1: Video is encoded at blocks of 16x16. If your resolution is not mod16, then the encoder will extend the video to mod16. So some pixels are added and coded, but will never be displayed (they'll be cropped away to restore your none-mod16 resolution at playback time). So in fact you waste some bitrate for pixels you never see. However with x264 the overhead for using a none-mod16 resolution is pretty small. So whenever possible use mod16. If it's not possible to use mod16, don't worry too much about it...

Q2: It depends on your source! If the source was anamorphic ("none-square pixels"), then the best way to keep as many details as possible is keeping the video at it's "native" resolution (that is: do NOT resize at all) and encode it anamorphic (as the source was) by setting up the proper PAR. If you want to encode an anamorphic source with "square pixels", it unavoidably has to be resized first. And this is the point where you'll loose details! Also if you "upscale" to restore the 1:1 PAR, you'll have more pixels to be coded and hence you'll need more bitrate. If you "downscale" to restore the 1:1 PAR, you'll loose a lot of details in this step. Finally if your source is NOT anamorphic, there's no sense in encoding it anamorphic...