D-Frames

[SirDavidGuy]

I've been working around with Mpeg-1, and I keep running into D-Frame's. I can't find any given definition for their format, or how to process them properly (other than encoding only the DC coefficient, which doesn't really tell me what I need to know).

[redeemer-dk]

AFAIK D-frames are frames ONLY consisting of DC coefficients. so my guess is that all the AC coefficients are zero which would set them to the same value as the DC coefficient. this could perhaps be the case in a total black block where the DC coefficient is only needed if the AC's are the same. try decoding like that and look at the product. let me know how it works :)

-Rune Svendsen

[mpucoder]

Right you are, and they are not permitted in MPEG-2.

SirDavidGuy: You will have better luck searching on "D-Pictures", the official name for them.
Get yourself a copy of ISO 11172-2 (available at Ookami's Programmer's Corner (http://ookami.videoxone.de/programmers.html) - the mpeg1.rar file, extract Mpgvideo.doc)
There, you'll see "DC-coded picture; D-picture: A picture that is coded using only information from itself. Of the DCT coefficients in the coded representation, only the DC-coefficients are present."
And at 2-D.6.6 is where they are explained.

[Valiant]

did i hear DCT? :)

perhaps you wanna know more about DCT...
i did a DCT iDCT with zigzagscan. if you want a fast DCT you've got to do it in integer and use 1. 2d DCT as 2x 1d-DCT on rows/columns and to do it even faster as DFT (Fourier) with lots of math you'll end up with Feig's algorithms.

see:
· 2D-DCT basics, visually explained:
http://rnvs.informatik.tu-chemnitz.de/~jan/MPEG/HTML/mpeg_tech.html
· 1D-DCT basics:
http://www.cs.cf.ac.uk/Dave/Multimedia/node231.html
· DCT analyzed– why 8x8 blocks, why DCT:
http://bs.hhi.de/~wiegand/transform_coding_02.pdf
· the JPEG Standard:
http://avs-www.informatik.uni-hamburg.de/teaching/ws-94.95/mm/vortraege/Kompr_verf/jpeg.htm
· very good explanation of the best known DCT algorithm (Arai, Agui, Nakajima, Feig):
http://rnvs.informatik.tu-chemnitz.de/~jan/MPEG/HTML/IDCT.html
· 3 different 2D-DCT implementations, not very accurate algroithms:
http://members.optushome.com.au/emikulic/gfx/dct/
· Berkeley MPEG IDCT improved with Feig (Source Code):
http://rnvs.informatik.tu-chemnitz.de/~jan/MPEG/HTML/idct_discussion/Index.html
· Arai, Agui, Nakajima with SIMD:
http://skal.planet-d.net/coding/dct.html

[redeemer-dk]

that's very interesting Valiant. Do you have "the JPEG standard" document in english?

[Valiant]

lol, sorry, didn't remember the language of the files.

no, i don't have the jpeg standard in English. google? perhaps you'd be more interested in jpeg 2000 (wavelet instead of DCT).

[SirDavidGuy]

Originally posted by mpucoder
Right you are, and they are not permitted in MPEG-2.

SirDavidGuy: You will have better luck searching on "D-Pictures", the official name for them.
Get yourself a copy of ISO 11172-2 (available at Ookami's Programmer's Corner (http://ookami.videoxone.de/programmers.html) - the mpeg1.rar file, extract Mpgvideo.doc)
There, you'll see "DC-coded picture; D-picture: A picture that is coded using only information from itself. Of the DCT coefficients in the coded representation, only the DC-coefficients are present."
And at 2-D.6.6 is where they are explained.


Ah, thank you. This is exactly what I needed.