I'm wondering what value of t would be best for lerping the results of the 2 bi-linear filters together. Say we're filling a space of 100 pixels, so we lerp between a sample from 128 texels and 64 texels. I'm thinking we would either lerp with t = (100-64)/64 = .5625 or t = log2(100) - log2(64) = log2(100/64) = .6438...
I feel like the second option would be more accurate, but the first option looks easier to compute
edit: ^ i guess the second option may not be as expensive since you have to compute log2 anyway to find which mipmap to use, and you could compute it once for the whole image, instead of per pixel
I'm wondering what value of t would be best for lerping the results of the 2 bi-linear filters together. Say we're filling a space of 100 pixels, so we lerp between a sample from 128 texels and 64 texels. I'm thinking we would either lerp with t = (100-64)/64 = .5625 or t = log2(100) - log2(64) = log2(100/64) = .6438... I feel like the second option would be more accurate, but the first option looks easier to compute
edit: ^ i guess the second option may not be as expensive since you have to compute log2 anyway to find which mipmap to use, and you could compute it once for the whole image, instead of per pixel