the idea is great but the implementation not so. Certainly there is a correlation between where the user points his mouse and where he looks - but it is pretty small. Especially when pictures and multimedia are involved people tend to move their mouse out of the way.
With a high enough quality webcam and flash (which will allow you to stream from the cam) it should be possible, not easy but possible, to get pictures of the eyes and track their movement. The cons of this approach are:
1) you need to get the users permission to view captures from their cam - major irritation and bad usability.
2) the software to anslyse the pictures is quite complex
3) the cam needs to be of a pretty good quality
4) It will be hard to determine the coordinates of the cam in relation to the monitor, though it is probably possible to approximate it, since you can assume that the user is staring at the screen his eye movement will be confined to a square of space somewhere in front of him (the monitor). If data is gathered for a while it should be possible to approximate where the screen is in relation to the cam.
As a mouse shover and a yc news reader, you would come under the demographic that blocks ads probably even uses lynx. So, "they" would safely count you out of their target audience. Ha!
With a high enough quality webcam and flash (which will allow you to stream from the cam) it should be possible, not easy but possible, to get pictures of the eyes and track their movement. The cons of this approach are:
1) you need to get the users permission to view captures from their cam - major irritation and bad usability.
2) the software to anslyse the pictures is quite complex
3) the cam needs to be of a pretty good quality
4) It will be hard to determine the coordinates of the cam in relation to the monitor, though it is probably possible to approximate it, since you can assume that the user is staring at the screen his eye movement will be confined to a square of space somewhere in front of him (the monitor). If data is gathered for a while it should be possible to approximate where the screen is in relation to the cam.