take this for example
0.00|0.00|0.00|0.00|0.00
------------------------
0.00|0.00|1.00|1.00|1.00
------------------------
1.00|1.00|1.00|1.00|1.00
if say (pseudo code)
scale = 2;
edgedirection.y = abs(right - middle);
edgedirection.x = abs(top - middle); //kinda like anisotropic bluring
edgerange = length(edgedirection);
color = tex2d(texturein, middlecoord);
for(i=1; i<=scale , i++){
color += tex2d(texturein, middlecoord + edgerange * i * pixelsize * edgedirection);
color += tex2d(texturein, middlecoord - edgerange * i * pixelsize * edgedirection);
}
color /= scale*2 + 1;
return color;
you would get (assume sampling in clamp mode)
0.00|0.00|0.00|0.00|0.00
------------------------
0.00|0.40|0.60|0.80|1.00
------------------------
1.00|1.00|1.00|1.00|1.00ofc the actual edge finding & weighting (using 1 here) would have more restrictions and depend on more terms, but you get the idea.
and since SMAA or FXAA are provided in framework, you can check their algorithm for details on how they sample and weight.
Topic Author
