You only want to make 384 x 288 AVI files, but want Higher Quality?
WITHOUT having to Deinterlace?

Well so do I, So here's something that may help

Oh before we get started, you will need Virtual Dub installed. If you don't have it, you can download a slightly old version from This Page, or you can go directly to the Virtual Dub Home Page, and obtain the latest version.

For me capturing at 768 x 576 has 3 MAIN limitations:

ONE Requires a faster CPU and more Disk space
TWO Requires more system resources (ie a faster CPU) for smooth playback. Especially related to A/V sync. I want to produce Div-X AVI files, and the system resources required to playback 768 x 576 Divx AVI files are too high.
THREE My Capture card can not capture Non Interlaced at 768 x 576 in Overlay or Preview mode.

So without spending more money on a faster PC with more resources, or higher spec capture card, what are my other options?. Well with some tests I've done, Capturing interlaced footage and then de-interlacing it, has been less than adequate. The quality is nowhere near what I'd be happy with for my final product. So that means limiting my captures to NON-interlaced frame sizes. After conducting some more tests, The highest NON-interlaced frame size my card can capture at is 768 x 288. Because the height doesn't go above 288 the footage stays non-interlaced.

But hangon, this is way WAY out of the 4:3 aspect ratio that most broadcasters use. (I'm in a PAL country where Full PAL resolution is recognized as being 768 x 576 (not 704 x 576)

So that means I must resize the frame back to 384 x 288. But which filter and settings should you use, and is there any difference than just capturing at 384 x 288?

Ok, so lets start making some comparisons. The footage I've chosen is quite bright and is of a monkey eating a cracker. The reason I chose this is because the monkey has lots of fur, and therefore any detail loss is easy to spot. The background unfortunately is out of focus, so we can't do anything about that, however you can still use it to see any artifacts or loss of detail.

Purpose of this page.

It seems ironic to me, that so many people go out of their way, or spend so much money just to capture Full resolution (such as PAL 768 x 576) to AVI. I see no real point in that, unless you intend to make MPEG DVD's or something, because as stated in the above 3 points, the resources required to playback such AVI files, put it out of reach for most PC owners. (especially if you intend to watch a 2 hour movie at 768 x 576, your system is almost flogged to death). Div-X is such a great codec that there is really no need to create such large framed AVI's. If you view any of the below sample avi files at full screen, they still look pretty sweet. Not to the standard of DVD, but an mpeg at similar bitrate and framesize would look considerably worse :)

So hopefully this page gives inspiration and hope to other encoders who suffer from the 3 main limitations I mentioned at the start of this page. You can still increase the quality of your captures without spending more money on new equipment or having to learn how to de-interlace.


First off A capture done at 384 x 288

Now a capture done at 768 x 288

Yuck. man that looks terrible, but don't worry, it will soon look *HEAPS* better. As you can see aspect ratio is right out the window. If you have a keen eye, you've probably noticed just how much more detail is visible in the larger second image. This detail is retained when we resize down.

Virtual Dub has some very good built-in Filters. We are going to use one of them now. The Resize Filter.

As you can see, we can enter any frame size we want, so for this test, I want the footage to be 384 x 288. Now the list of options is slightly longer than what is displayed, however the options I don't display are quite useless when it comes to quality. These options are Nearest Neighbor, BiLinear and BiCubic. They all introduce extreme amounts of noise in the frame (sample images are displayed at the bottom of this page if you are curious). The first option shown is Precise BiLinear. Now I find Precise BiLinear makes the image too soft, (lacks sharpness) so the First real usable option is Precise BiCubic (A=0.75), any of the last three files all produce acceptable results and it's quite difficult to spot any real difference between the three, other than using (A=0.60) produces files which have a slightly smaller size and (A=1.00) produces slightly larger files. Filesizes are shown later in this document.


Now it maybe a little hard to display the precise differences in each test when the image is this small, so I've also produced 1024 x 768 enlargements (captures during playback at my desktop resolution of 1024 x 768) to make the comparison easier. Another main reason why I created the 1024 x 768 samples, is because when I watch anything I capture, I prefer to view it in fullscreen, which is at my desktop resolution of 1024 x 768. So I wanted to see not just the frame size quality, but also the playback frame size quality. Here is a side by side comparison to give you an example of the potential benefits.

COMPARiNG 384 x 288 versus 768 x 288 versus 768 x 576
Capture at 384 x 288

Loss of sharpness and fine detail is quite obvious.
Capture at 768 x 576

Just for a comparison.
Capture at 768 x 288

Resized down to 384 x 288 using Precise BiCubic (A=1.00).

You can clearly see the additional sharpness gained over capturing just at 384 x 288, but there is a reduction in quality compared to a full resolution capture at 768 x 576. However this 768 x 288 has been bicubic resized down to 384 x 288, then captured at playback resolution of 1024 x 768. If I do a more direct comparison of bicubic resizing 768 x 288 upto 768 x 576, then capturing at playback resolution of 1024 x 768, you will see the variation between the two is not as great.
COMPARiNG 768 x 288 versus 768 x 576
Capture at 768 x 288

Resized down to 384 x 288 using Precise BiCubic (A=1.00)
Capture at 768 x 576
Capture at 768 x 288

Resized UP to 768 x 576 using Precise BiCubic (A=1.00).

You can see an additional amount of sharpness and fine detail is retained when resizing up compared to resizing down.
Comparison image captured at desktop resolution. ( 1024 x 768 ) Only partial frame shown

I capture to Huffyuv lossless codec first, and resized down, but still stored using Huffyuv lossless codec. So the only variation you should see is what was done by the resizing filter.

No other filters were used on the footage. Only the resize filter where appropriate. All footage and frames are untouched. Please take into consideration that these images are JPEG compressed and that may also remove fine detail, but BMP or TIF images are just way to large to use on a web page.

HUFFYUV Frame Comparison Of each Resizing Method
384 x 288
768 x 288

768 x 576

Sample Video Footage

If you would prefer to see first hand what those settings look like, here are 2 second samples from each setting. All of the above still frames were taken from the below footage. The footage was recorded from a PAL analog input using the tuner on a Pinnacle PCTV Rave card.

DivX version 5.00 or Higher is required to view the below AVI files

Captured at 384 x 288
No Resizing, 384 x 288 Capture 337,920 bytes
Captured at 768 x 288
Precise Bicubic (A=0.60) 487,424 bytes
Precise Bicubic (A=0.75) 497,664 bytes
Precise Bicubic (A=1.00) 518,144 bytes
Is *OK* but not my Preference.
Precise Bilinear 444,416 bytes
These Settings all Look Terrible and are not worth using
Nearest Neighbor 749,568 bytes
Bilinear 749,568 bytes
Bicubic 749,568 bytes


To show the actual filesize difference between two filter options, I did a real life length capture . The movie duration was 1 hour 52mins and 21 seconds. Post-Encoded with Div-X 5.00 using 1-Pass Quality Based encoding set at 93%. Captured at 768 x 288 and resized to 384 x 288

Precise Bicubic (A=1.00) 463,593,472 bytes 3.93 MB per minute, 84.81:1
Precise Bicubic (A=0.60) 447,244,288 bytes 3.80 MB per minute, 87.27:1


FRAME Quality - 384 x 288 capture
The 384 x 288 capture isn't bad, but when compared to the Precise Bicubic resizes, it looks slightly soft (especially when you compare them on a frame level). Slight fine detail is also lost. Overall not to bad but can be improved.

FRAME Quality - 768 x 288 captures
Well, there's no real comparison In My Humble Opinion. Comparing the Huff and the Divx frame, capturing at this frame size and resizing down does make a small improvement to the sharpness and fine detail component of the frame. Advantages are also gained in High contrast areas.

ADVANTAGES Of - 768 x 288 capturing
It can easily improve the fine detail and sharpness of each frame.
Advantages are also gained in High contrast areas
Potentially allows you the highest quality non-interlaced captures your card can do.

DIS-ADVANTAGES Of - 768 x 288 capturing
It will be slower to process during Post-Encoding.
Requires double the disk space during capturing.
Produces slightly larger post-encoded files .
Potential to produce staircase artefacts if the frame contains straight lines. This would have to be the largest disadvantage.

I think it will be upto each individual to determine just how important the content quality is for each capture they make. I know I won't be capturing at 768 x 288 for everything I capture, because some of my captures, I only watch once and then delete, so this is would be a huge waste of disk space and post-encoding time. However for DVD recording, where the source input quality is very high, it is certainly worth doing to ensure that fine detail is retained.

If you found this page helpful, you may also wish to check out my other Frame Size Comparison Page where I go into more detail using other resolutions (such as 640 x 480, 384 x 576 and 768 x 576).

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

This Page was:
Created on: 9th February 2oo3
Last updated on: 29th September 2oo4
(Added 768 x 576 image comparison)

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