I've tryed this way first. But for good results you have do use this algorithm for each Pixel.

pixel = pixel_for_vec(vector(PixelColour2[0],PixelColour[1],0.7*PixelColour[1]+0.3*PixelColour[0]),50,format);

PixelColour1[0] = blue value first view
PixelColour1[1] = green value first view
PixelColour2[0] = blue value second view
PixelColour2[1] = green value second view

I tryed views with coloured objects in front of the camera, too. But I never get the same results like "filtering" the colour information with this formula.

For each pixel you have to take the blue and green value from one view and calculate the red value from the blue and green value of the other view. And so you get an partly colourd anaglyphic picture containing all colours except red.