XviD.Ru Bitrate and quality. Why less is not worse


Яндекс.Метрика Рейтинг@Mail.ru правильный HTML5 правильный CSS
В рисунках детей - неиссякаемое добро, любовь и свет!
Russian version here

Some people think "more bitrate - more quality". There are some examples. They demonstrate some important methods do decrease needed bitrate with increasing result quality.

VHS and DVD rips. Compare sizes

But compare the most obvious situation: VHS cassette of bad quality and DVD of good quality.

Try to pack both streams resized to identical width*height dimensions with fixed quantizer (in DivX, XviD), better 1-pass with quantizer 2.

Compare sizes of result files. Note, that in the MOST cases you get less size for DVD-rip version, and sometimes this difference is incredibly great (3-4 times). What does it mean?

It means, that the most part of bitrate is used for interlacing and other artifacts of VHS original, garbage from analog TV signal. NOT really from original picture itself. Less garbage - less additional bitrate needed. Difference from one cassette to another may be great.

So, if you see DVD-rip of 1CD size, it may be and in most cases is better than VHS-rip of 2CD size. If you see two VHS-rips of the very different sizes (even 1CD - 3CD), smaller rip MAY BE better!

If VHS rip isn`t properly telecided/deinterlaced - it will be worse or/and larger. See next part of this text.

Why we must do all possible inverse telecine?

NTSC 2-3 stream contains duplicated information. If we interpret it as progressive frames and pack with any MPEG algorythm, duplicated fields are treatened as parts of different pictures. Because MPEG loses some information in every frame - these pictures become MORE different, than original ones. Nore difference - more bitrate needed.

NTSC oriinal. Reduce size. Packing with fixed quantizer 2.

Pictures to the right show small part of frame enlarged.

Pack without IVTC process. Keep 29.97 fps and all interlaced frames as is. Note, that interlacing artifacts are "moired" when resizing. Example lines in AviSynth:

BicubicResize(640,340,0,0.1,8,8,704,458)


321k size - result video file.

Pack without IVTC process, but Deinterlace it with blending. Reduce size. Pack it with fixed quantizer 2 and see size.Example lines in AviSynth:

FieldDeinterlace()
BicubicResize(640,340,0,0.1,8,8,704,458)


296k size - result video file.

Take the same material and DO IVTC process properly, cleaning interlace artifacts completely. Example lines in AviSynth:

Telecide(post=false,guide=2,blend=false)
Decimate(cycle=5)
BicubicResize(640,340,0,0.1,8,8,704,458)


254k size - result video file.

Compare sizes. Is result now obvious for you? Third variant is smallest and the best for viewing.

What else? SpatialSoften()

Effective way to clean some trivial garbage and to decrease needed bitrate is to use SpatialSoften() and TemporalSoften() filters in AviSynth.
  1. This setting is possible for normal film:

    SpatialSoften(4,4,16)


    235k size - result video file.
  2. This setting is VERY useful for cartoons:

    SpatialSoften(8,8,32)


    223k size - result video file.

Remove duplicates from cartoons

Using Dup() plugin for AviSynth help you in packing cartoon movies, if they have movement not in every frame. For example, many anime movies have only 15fps for characters, hungary cartoon "Vook" has full every-frame movement only for background, not in all scenes. Characters are moving only in every second frame.
© 1999-2015 Sergej Qkowlew. Разрешается использование текста, используемых в нём графических файлов, яваскриптов, файлов с табличными данными и элементов оформления полностью или частично в любых целях, КРОМЕ как для дезинформации кого-либо относительно материалов, произведений искусства, проектов, лиц, организаций, событий. Ссылку на источник давать разрешается.