I'd definitely be interested in an explanation of more modern graphics techniques such as this one, since it seems like a lot of the options in standard video game menus (such as filtering, type/quality of antialiasing, etc.) have made a lot more sense and have been interesting learning about taking this course.
keenan
Slide 19 of these slides has a pretty good illustration: if a pixel is stretched out vertically in texture space (for instance), you want to take an average over a tall skinny rectangle. And you can easily do that by building on top of an existing MIP map implementation: rather than compute a single quantity $L$ that determines the MIP map level, compute an independent width/height for the projected pixel. Use the smaller of the two to choose the MIP map level. Then average multiple MIP map samples from that level to get an estimate of the average under the pixel. E.g., in the example above (and on the slide) you want two trilinear MIP mapped samples that are offset in the vertical direction.
I'd definitely be interested in an explanation of more modern graphics techniques such as this one, since it seems like a lot of the options in standard video game menus (such as filtering, type/quality of antialiasing, etc.) have made a lot more sense and have been interesting learning about taking this course.
Slide 19 of these slides has a pretty good illustration: if a pixel is stretched out vertically in texture space (for instance), you want to take an average over a tall skinny rectangle. And you can easily do that by building on top of an existing MIP map implementation: rather than compute a single quantity $L$ that determines the MIP map level, compute an independent width/height for the projected pixel. Use the smaller of the two to choose the MIP map level. Then average multiple MIP map samples from that level to get an estimate of the average under the pixel. E.g., in the example above (and on the slide) you want two trilinear MIP mapped samples that are offset in the vertical direction.