hpdz.net

Please be patient! There are 5 uncompressed BMP images being downloaded!

This page shows examples of some techniques for anti-aliasing. Farther down is an explanation of how each test image was made. There are also three videos demonstrating methods A, B, and C.

Sorry, but I haven't had time to write a good explanation of any of this. Hopefully those who are already familiar with the subject will find these examples useful. There are more powerful methods of achieving similar effects that I have my research department actively investigating.

A high-resolution, extreme oversampled version of this image is way down at the bottom.

These are the final frame of Centanimus, with a size vertically of 6.6E-102.

Which one do you think looks best? Worst?

Image A	Image B
Image C	Image D
Image E

Here are the links to the UNCOMPRESSED animations ending on the above images. These are uncompressed 320x240 AVI files and are about 83 MB each.
MethodA
MethodB
MethodC

I have made a 500 Kbps MP4 file with all three methods demonstrated. It is only 4.5 MB, which is about 1.8% of the original uncompressed file sizes, but it still demonstrates each effect quite clearly.

I have also made a 500 Kbps WMV file (2-pass CBR) of this for those having trouble with the MPEG-4 video file. It is significantly smaller at only 3.2 MB and also has fewer visible compression artifacts, but seems to have some frames dropped near the end of each segment.

921KB 640x480 BMP
This image is a high-resolution rendering with a median filter and 16X stochastic supersampling. Sure is pr'ty, but at 5.74 hours of computational time to render on my 3.2 GHz Pentium-IV, it is out of my league for animations (even if I used the quad-Core2 system, it would still be impractical).

Descriptions of Techniques

Method A

Plain image with no modifications.

Method B

1X stochastic sampling. The same size sampling grid was used, but each point was jittered by a random amount. This has the effect of spreading out the aliased energy over the entire spatial frequency spectrum, rather than concentrating it at a few low frequencies. Moire is essentially gone, but it is replaced by dramatically increased noise in the areas where the image is undersampled.

Method C

Stochastic supersampling on a 2x2 grid with median filtering. Each image pixel is the median of four supersampled points taken on a jittered 2x2 supersampling grid.

Method D

Stochastic supersampling on a 3x3 grid with median filtering. Each image pixel is the median of four supersampled points taken on a jittered 3x3 supersampling grid.

Method E

Non-jittered supersampling on a 3x3 cross with median filtering. This greatly reduces speckle noise, which is one effect of aliasing, but has much less of an effect on moire. That is because moire is a consequence of the spatial frequency of the raw data being higher than that of the pixel grid, and filtering like this doesn't reduce the spatial bandwidth much. The filter window would have to be wider.

Demonstration of Median Filter to Remove Speckle Noise

This is a phenomenon related to aliasing that has a different visual appearance than moire. This is pretty self-evident and doesn't need much explanation. Filtering correctly removes it at the expense of horrific increases in rendering time to do the extra oversampling. There is also some subtle moire evident, much more so in the unfiltered image. The filtered image below was generated by applying a median filter on a 3x3 supersampling grid for each pixel.

These are zooms at a size of 1.2e-15.