Looking at the walls, a number of the tris seem to be inconsistently colored in a way that seems to "mess with" the gradient. The archway on the right side, for example, seems to have a bunch of green/orange "spikes" as alternating tris are colored differently. Is this intentional? Why is it that the mipmap level in use doesn't scale gradually between adjacent tris like this?
yauwilliam69
I am not sure - I think the tris has a higher level than the rest around them. The tris have a geometry where there are more folds than the rather smooth surfaces around them. This extra folding causes the du/dx in the texture space to increase even more than those around them, with a similar perspective angle. Basically, the tris has to represent more texture pixels to give that geometry, causing mipmap to raise its level to avoid aliasing.
keeratsingh2002
how does trilinear filtering bridge the transition between these Mipmap levels to create this experience? How does it select the appropriate Mipmap level to avoid noticeable changes in texture detail as the viewpoint moves?
pranavkolluri
My intuitive way of understanding the mipmap layer system is that the more the "view" of the texture changes, the higher the mipmap layer to use since more smoothing would be needed. Would it be correct to then understand the mipmap levels on this map to some degree show the amount of geometric change (I guess the geometric complexity from the camera view)?
Jamie-Walton
I found it helpful to think of the process like creating a multi-colored gradient. First, you pick discrete colors and place them in the order you want. This will give the effect of color changes, but it's still choppy. To fix it, you interpolate between each color to make smooth transitions and create one cohesive gradient. In this case, the mipmap levels are like the discrete colors. The algorithm "chooses" colors with the equation from the "Computing Mipmap Level D" slides (I'm guessing that the number of levels is up to the programmer to balance accuracy to the texture map and efficiency). The trilinear filtering is the interpolation between each level, which creates the smooth transition. The more mipmap levels, the less heavily the graphic relies on filtering to fill in the gaps.
Looking at the walls, a number of the tris seem to be inconsistently colored in a way that seems to "mess with" the gradient. The archway on the right side, for example, seems to have a bunch of green/orange "spikes" as alternating tris are colored differently. Is this intentional? Why is it that the mipmap level in use doesn't scale gradually between adjacent tris like this?
I am not sure - I think the tris has a higher level than the rest around them. The tris have a geometry where there are more folds than the rather smooth surfaces around them. This extra folding causes the du/dx in the texture space to increase even more than those around them, with a similar perspective angle. Basically, the tris has to represent more texture pixels to give that geometry, causing mipmap to raise its level to avoid aliasing.
how does trilinear filtering bridge the transition between these Mipmap levels to create this experience? How does it select the appropriate Mipmap level to avoid noticeable changes in texture detail as the viewpoint moves?
My intuitive way of understanding the mipmap layer system is that the more the "view" of the texture changes, the higher the mipmap layer to use since more smoothing would be needed. Would it be correct to then understand the mipmap levels on this map to some degree show the amount of geometric change (I guess the geometric complexity from the camera view)?
I found it helpful to think of the process like creating a multi-colored gradient. First, you pick discrete colors and place them in the order you want. This will give the effect of color changes, but it's still choppy. To fix it, you interpolate between each color to make smooth transitions and create one cohesive gradient. In this case, the mipmap levels are like the discrete colors. The algorithm "chooses" colors with the equation from the "Computing Mipmap Level D" slides (I'm guessing that the number of levels is up to the programmer to balance accuracy to the texture map and efficiency). The trilinear filtering is the interpolation between each level, which creates the smooth transition. The more mipmap levels, the less heavily the graphic relies on filtering to fill in the gaps.