Ok, it's time to tackle the mosquito noise problem.

I think we should start by summarizing what we know so far, and then we can attempt to work out a way to fix it.

Here's my thoughts:

It occurs at boundaries between high-frequency and low-frequency areas.

It is more noticible on the low-frequency side of the boundary.

Anyone else notice any other characteristics we can take into account to target it with a filter?

Hi SansGrip,

One of the things I notice, I believe I mentioned before, is the effect on a object when it is far away in the picture. When a object is far, and it's size is close to the actual pixels of the encoded resoution, it creates an artifact on movement. Probably when the object moves from a even to an odd pixel, you see this effect. On higher resolutions, this is not so visible. The only way I can describe this is sort of a "Moire Pattern", that happens on graphics. I wonder if the old "Indian" black and white test pattern would be any good to test your filters That pattern clearly shows the Moire Pattern's effect. Of course, this is a still image test pattern, so I'm not sure if it would be of any help.

-kwag

Hmmm. There might not be much we can do about very distant objects. But we can certainly try to reduce mosquito noise in high-contrast areas.

In order to test I'm going to need to encode with settings guaranteed to produce mosquito noise. Any recommendations?