x264 manual (spanish) fixed

[jjseth]

New manual for x264 configuration and use in spanish.
Section "Guías" at SethNet.tk

[IgorC]

In one part it says : H264 vs. MPEG-4 . But H.264 ia MPEG-4 (part 10) too. It will be better : H.264 (MPEG 4 part-10 or AVC ) vs MPEG-4 ASP. I know that you understand difference but if some newbie will read it.

[akupenguin]

(Forgive me if I misunderstand something; my spanish isn't perfect.)

Cuanto más se dividiesen los bloques más precisa sería la estimación del movimiento y MPEG-4 llegó al máximo con la aplicación de las q-pel, o lo que es lo mismo la división de bloques hasta el quarter-pixel (1/4 de píxel).
MPEG-4 AVC vuelve a ir más allá llegando a dividir los bloques hasta 1/16 de píxel.

You're confusing block size with motion vector precision. Both H.264 and ASP use qpel. ASP divides the frame into blocks of a minimum size of 8x8 pixels, while H.264 allows down to 4x4.

¿qué sucede cuando no hay movimiento?. Para la representación de paredes y otros objetos estáticos no se puede hacer una predicción normal de cada píxel en el siguiente frame ya que al fin y al cabo los píxeles no van a variar.
H264 analiza los píxeles vecinos para intentar encontrar las posibles modificaciones y compararlas con el píxel actual de manera que la representación inmóvil sea más sencilla.

There is no difference at all between H.264 and ASP regarding prediction of static areas. Both use the perfect prediction: copy the pixels from the previous frame.
Maybe you meant to include some mention of Intra prediction instead?

Mientras MPEG-4 trabaja, por decirlo de alguna manera, con decimales donde a la larga tras muchas operaciones se termina por "perder" números; en H264 se trabaja con números enteros por lo que no se pierde información. De esta manera se evitan macrobloques, "efectos mosquito", etc.

It's true that H.264 uses integer arithmatic while ASP uses real-valued DCT and requires the codec to make approximations. But that has no significant effect on quality in normal encodes. Its purpose is only to allow long GOPs (e.g. more than 300 frames) without the accumulation of errors that would be caused if the decoder makes different approximations than the encoder. (And because it's better for the standard to define exactly what a bitstream means, rather than leaving the details up to the implementation.) The only times I have ever seen noticable artifacts due to DCT drift were in DivX3.

H264 incorpora un filtro de Deblocking o lo que es lo mismo un filtrado de reducción de macrobloques que actua sobre los macrobloques y los bloques de 1/16 de píxel.

That's "macroblocks and 4x4 blocks".

[jjseth]

@IgorC:
Ok, I'll specify the name of mpeg-4 version.

@akupenguin:
Thanks for the corrections. I will edit the manual.

You're confusing block size with motion vector precision. Both H.264 and ASP use qpel. ASP divides the frame into blocks of a minimum size of 8x8 pixels, while H.264 allows down to 4x4

Yes, in x264 "Macroblock Partition" section I wrote:
-----------
...sea cual sea el nivel elegido de precisión de este códec emplea siempre las conocidas Q-pel.
-----------
I'll edit the manual for being more precise.

There is no difference at all between H.264 and ASP regarding prediction of static areas. Both use the perfect prediction: copy the pixels from the previous frame.
Maybe you meant to include some mention of Intra prediction instead?

Yes, I try to explain the x264 use of intra prediction but I'll improve this section (or I'll try it xD).

It's true that H.264 uses integer arithmatic while ASP uses real-valued DCT and requires the codec to make approximations. But that has no significant effect on quality in normal encodes. Its purpose is only to allow long GOPs (e.g. more than 300 frames) without the accumulation of errors that would be caused if the decoder makes different approximations than the encoder. (And because it's better for the standard to define exactly what a bitstream means, rather than leaving the details up to the implementation.) The only times I have ever seen noticable artifacts due to DCT drift were in DivX3.

Thanks for explain this part that was confusing to me.

That's "macroblocks and 4x4 blocks"

One (stupid?) question:
16x16 block = 1
8x8 block = 1/4 (q-pel)
4x4 block = 1/16?
Or I'm wrong since the beginning of block's size? xD
Maybe likewise I'll change 1/4, 1/16 by block's size.

My english is worse than your spanish

[sysKin]

Originally posted by jjseth
One (stupid?) question:
16x16 block = 1
8x8 block = 1/4 (q-pel)
4x4 block = 1/16?
Or I'm wrong since the beginning of block's size? xD
Maybe likewise I'll change 1/4, 1/16 by block's size.

You are completely confusing precision of the vector (4th pel, halfpel...) and the size of the block described by this vector (16x16 pixels big, 8x8 pixels big...)

[jjseth]

Originally posted by sysKin
You are completely confusing precision of the vector (4th pel, halfpel...) and the size of the block described by this vector (16x16 pixels big, 8x8 pixels big...)

Yes, like akupenguin said I "confusing block size with motion vector precision".
I continuing learning .
Thanks sysKin

[jjseth]

Manual fixed (or I want to believe it... )