WHAT FRAME FORMAT YiELDS THE HiGHEST QUALiTY.
But what are the Advantages / Dis-advantages associated with that frame format


WHAT iS FRAME FORMAT ?

Pretty simple. Imagine each frame in an AVI file is a single picture, just like a normal image/picture, such as .TGA .BMP .JPG .PCX or .PNG for example. Now each image format has its own benefits. For example TGA (Targa) is a very high quality format but it is also uses lots of disk space, whereas .PNG is a very high compression format. So the same goes for AVI files. You get to select what level of quality and compression you want for each frame in an avi file. Using PAL, that equates to 25 images per second (25 fps). So the format you select can play a very large part in two ways.

FIRST WAY
It will determine the quality (level of detail) that each frame stores in the avi file. The result is you get good detail and colour. High Quality footage.

SECOND WAY
It will determine the size that each frame uses in the avi file. (determines the filesize of your captures). The higher the quality you use the more disk space you will need during capturing.

Everything is a compromise, Quality versus Quantity. You can, unfortunately, never have both.

This is the selection screen in AVI_IO. All capture programs that are worth anything should have a menu option somewhere which allows you to select the Frame Format you wish to use. The range of options you get could be related to the type of tv-card / capture card you have, or what codecs you have installed. They could even be influenced by the type and quality of drivers you are using to control your tv-card.

FiLESiZE RESULTS

OK.... FIRST thing to look at is the relationship between frame format and filesize?.. Is there any?....

Test Capture from my Pinnacle PCTV Rave PCI Card, using S-Video input from 4:3 25fps PAL DVD-ROM (Region 4). Capture duration is for 1 minute (1,500 frames) with NO audio and capturing to RAW uncompressed AVI. (no compression).
Each capture was using Identical footage and captured at 384 x 576 frame size. See This Page for reasons why I chose this resolution.


FRAME FORMAT FiLESiZE ADDiTiONAL iNFORMATiON
32 bit RGB 1,327,898,056 bytes - none -
24 bit RGB 1,001,434,856 bytes - none -
16 bit RGB 668,211,832 bytes Produced no image on my PCTV Card. (all frames black)
15 bit RGB 669,658,856 bytes - none -
YUY2 668,211,832 bytes - none -
UYVY 669,211,832 bytes - none -
BTYUV 503,434,424 bytes Frame requires Vertical Inversion
YUV9 379,354,856 bytes Frame colour look worse than all other formats.
YUV12 503,434,724 bytes - none -


As you can see, for a single minute of capturing, the Frame Format that you select plays a HUGE part in the resulting filesize.


FRAME QUALiTY

Now how does each frame look. For example: Does the 62% additional filesize requirement for 32bit RGB make it better than BTYUV ?

View of Each frame direct from the RAW capture, They have NOT be de-interlaced.
             


DiSCUSSiON

I selected the footage for this test, because it maintained a static background, but still had reasonably fast moving foreground movement. The man (Craig Charles) is in the middle of laughing so the movement of his body is quite fast.

You can clearly see which frame formats handle static parts of the frame differently to moving parts. The 32bit RGB frame for example has no field separation in the static areas, and has only a small bit of field separation in the motion areas. However YUY2 handles it completely different. Field separation is everywhere.

So Why is this important I hear you ask?.. Well when you goto deinterlace this footage during post-processing you have to use different settings in order to deinterlace the best. If you use the same filter and settings to deinterlace 32bit RGB as you do YUY2, what will happen is the 32bit RGB frame will NOW have field separation in the static areas (where it wasn't before. You will add artefacts to the footage). I cover this further down in this document.


HOW TO DE-iNTERLACE, AND THE DiFFERENCES

As I mentioned above, each frame format has to be handled differently. There are numerous filters available which can de-interlace. The one I like is listed in the table below. There is another one by Gunnar Thalin (Deinterlace smooth), and I cover that in more detail on Another Web Page

The main things you have to remember, is that in interlaced footage, each frame is comprised of two separate images. These images are called fields. The way the fields interact with each other determines how well the footage will look for example, a very common way to de-interlace is to "blend" the two fields together. This I think looks terrible and removes sharpness from the footage. I prefer to Interpolate the fields. The other things you can do to fine tune the filters is to determine the order in which the two fields are address (odd and even), you can reverse this order and sometimes this is all that is needed. You can also specify that the filters, only compare the differences between the frames, or between the fields.. Each setting will either provide you with a positive or a negative. So experimentation is the key to finding the right setting for your captures.

FiLTER NAME AUTHOR VERSiON SETTiNGS USED
Virtual Dub internal Resize Avery Lee 1.5.3 (build 16250) bicubic A=-1.00
Smart Deinterlace Donald Graft 2.7 beta 2 Frame & Field differencing, Phase Shift
This is the configuration panel for the Smart De-Interlace filter, and selected are the settings I would use on the YUY2 frame footage. For the RGB frame footage, I would unselect the Phase Shift option to get the best results.

A really cool feature of this filter is the Show Motion Areas Only. This is great for fine tuning the Motion Threshold. It's also good for showing how noisy your footage is. If you select this, any portions of the frame that contain no movement, stay GREY. Any moving areas show through as the normal frame. So if your frame appears to have no motion, but when you run this, and there is not much grey showing, then it's probably noise. (noise is never static. heheh oops by static I mean void of movement, not poor reception..).

As an example with this footage, the 32bit, 24bit RGB, BTYUV, YUY2 and UYVY I was able to set the Motion Threshold to 5, due to the frame clarity, however 15bit RGB I had to set at 10 and the others at 8. (lower value is better). Even though I was capturing the exact same footage and from a DVD, each frame format handles it slightly differently, causing the need to fine tune the deinterlace filter.


FRAME QUALiTY After De-Interlacing

You've seen, how much disk space each frame format uses, and you've seen the raw capture footage, so now the real test, once the footage has been de-interlaced, which one wins. Is the field separation in static and non-static areas important in preserving image quality?. What about that huge filesize increase, did that yield any additional advantage.?. Have a look..

View of Each frame after DE-Interlacing, All Frames were taken from the post-processing files. I post processed to Huffyuv to keep the comparisons lossless.
             


DiSCUSSiON

Well it's quite a booooooooring slideshow isn't it?.. Did you spot the massive differences?.. Nope. They're just not there. The BTYUV and the YUV9 showed clear signs of bleeding with the red (look at the letters RED D) Where these red letters meet the white they leave a clear red haze, which the other formats don't have. Also YUV9 had massive artefacts during transitional frames (these are frames which have fields from two different scenes). No other frame format suffered from this.

The Best place on the above slideshow to notice any real difference between each frame format is the right side of Craig's Forehead (where the image is slightly darker)

There is no doubt that using the Show Motion Areas Only feature of the Deinterlace Filter, clearly helps identify which frame format yields the cleanest image. 32bit RGB and 24bit RGB where clear winners here, but YUY2 was almost equal.


CONCLUSiON

Taking everything into account. Filesize, CPU usage, deinterlacing ability, image quality and equally important codec support, I would definitely have to say YUY2 is my choice..It is also the optimum frame format to use for Huffyuv Lossless codec

If you prefer to not use Huffyuv, but instead like MJPEG based codecs which usually yield a slightly higher compression ratio than Huffyuv, then you're pretty limited to 24bit RGB. However as the above slide show displays, there is no additional quality gained by using this on my TV-Card. So it's only benefit is in allowing you the option to use RGB based codecs.

ADVANTAGES vs DiS-ADVANTAGES

FORMAT ADVANTAGES DiS-ADVANTAGES
32 bit RGB
  • Ability to capture maximum detail (is dependant on your tv-card though)
  • Fairly easy to de-interlace as any non-motion sections show no field separation
  • Highest CPU usage from all Frame Formats
  • Is *NOT* supported by all codecs. (I only have options for Cinepak by Radius, Microsoft Video 1, DivX 5.0, Microsoft MPEG-4 and Intel Indeo Video 4.5
  • Drastically increases the disk space requirements.
  • 24 bit RGB
  • Same as 32 bit RGB.
  • It *IS* supported by more codecs than 32 bit RGB. (I now have the additional codecs to select from: Inteo Indeo 3.2, Indeo Video 5.1, UYVY, Huffyuv, MJPEG, PCLEPIXL and MiroVIdeo 32XL
  • Second highest CPU usage from all Frame Formats
  • Second largest usage of disk space.
  • 16 bit RGB
  • - none -
  • Third highest CPU usage of all the Frame formats.
  • Is *NOT* supported by all codecs. (I have less options than 32 bit RGB. My list now is only: Microsoft MPEG-4, PCLEPIXL, MiroVIDEO-XL)
  • May not be supported by all Tv-Cards. My Pinnacle PCTV Card for example does not function well with it.
  • 15 bit RGB
  • Same as 32 bit RGB
  • Slight loss of colour vividness when compared to 32 bit RGB and 24 bit RGB
  • Is *NOT* supported by all codecs. (Huffyuv and MJPEG are not available)
  • YUY2
  • Uses less CPU time than any RGB format
  • Fairly easy to de-interlace, and interpolates well
  • Does not suffer from the colour loss like 15 bit RGB. Frame quality appears equal to 24bit RGB (once deinterlaced)
  • Does NOT work with MJPEG codecs (or any codec that requires RGB)
  • Frame shows clear signs of field separation, in motion and static areas
  • A little bit more difficult to deinterlace than RGB formats.
  • UYVY
  • Appears equal to YUY2
  • Frame shows clear signs of field separation, in motion and static areas
  • Less codec choices than YUY2 though
  • BTYUV
  • Frame quality appears very equal to 24 bit RGB
  • NO usable codecs to choose from. (only microsoft MPEG-4)
  • Image appears upside down on my Pinnacle PCTV Card, so vertical inversion is required during post-processing
  • After de-interlacing clear signs of colour bleeding is evident.
  • YUV 9
  • Used the least amount of disk space
  • Had the worst image quality (especially relating to colours)
  • NO usable codecs to choose from. (only microsoft MPEG-4)
  • Very poor ability when dealing with transitional frames. (frames that have two fields from different scenes).
  • YUV 12
  • Frame quality is similar to UYVY, with slight detail reductions in certain areas
  • Frame shows clear signs of field separation, in motion and static areas
  • Only codec to choose is Div-X and Microsoft MPEG-4


  • NOTES

    I'd like to point out that I don't have every codec installed on my pc, so I can only state which codecs I have installed that operate at the above mentioned frame image formats. If you have information on other codecs that are worth mentioning, just let me know.

    It is also worth mentioning, that some tv-cards can not operate in RGB format at all, and on these it's typical for the capture program to do the conversion from say YUY2 or UYVY to RGB. This intern requires more cpu time during capturing and wastes disk space for absolutely ZERO image quality improvements. So on these cards you would only have to use RGB frame formats, if you wish to capture in an RGB based codec (such as MJPEG).


    I Hope you found this page helpful. Have fun experimenting with your capture card and settings.

    If the Javascript slideshow doesn't work for you, or If you would prefer to view each image separately you can from the below links. Or right click and select "save target as"

    Raw Frame (interlaced) Post Processed (de-interlaced)
    32 bit RGB
    24 bit RGB
    15 bit RGB
    YUY2
    UYVY
    BTYUV
    YUV9
    YUV12
    32 bit RGB
    24 bit RGB
    15 bit RGB
    YUY2
    UYVY
    BTYUV
    YUV9
    YUV12

    This Web Site, requires
    Microsoft Internet Explorer with JavaScript And Frames Support
    Best viewed in 1024 x 768 resolution

    This Page was
    Created on: 3rd August 2oo3
    Last Updated on: 19th September 2oo4

    (C) Narler
    If you didn't get to this page from my Main Page, Click here