I mean, how should we get the detail information for upsampling?
Like the skins may have more detailed shaped and color in small scale, should we just subdivision it?
Could texture mapping be applied to upsampling such that when filling up the meshes we upsampled, we can use the original mesh as textures?
How can interpolation helps increase resolution when the "true" details have been lost?
Would bilinear filtering not be downsampling, as we are increasing the resolution and then going down to a lower one by taking averages?