Lecture 5: Texture (61)
jamespear

I'm still a little confused by the motivation behind mipmap. I understand that each map can save space and we may not need high resolution for certain textures, but what do we do with each of these maps and how do we combine them into a program?

brandonlouie

@jamespear I think the previous 3 slides doing a good job at highlighting the motivation behind mipmap. We can see that the textures closer to the camera look better at lower mipmap levels, but the textures farther from the camera look not as great due to aliasing. However, if we increase the mipmap levels, the opposite is true; the farther textures look better while the closer textures look blurry. I believe that if the mipmaps are used effectively, I think we can get the best of both worlds, with both closer and farther textures looking more clear. I'm not sure how you'd combine them into a program, but evidently it's possible accordingly to slide 66 (Visualization of Mipmap Level)

jinweiwong

When I used to do iOS app development on Xcode, whenever I imported an image file, it would automatically Mipmap-ify it, and now I know what it's actually doing!

anavmehta12

My intuition behind the Mipmaps is that as we move farther from a scene we stop noticing or being able to see the fine details of that scene so we can you higher level mipmaps with lower resolution without anyone noticing. However as we move closer and closer and magnify the scene we need that detail so we need to use lower mipmap levels.

Songbird94

Are we using mipmap to reduce resolution for textures from a distance and in doing so they would actually look better?

cvankeuren

From my understanding, mipmap is utilized when we're attempting to map a large texture to a much smaller surface area. Another way that it's helpful is that it allows us to pre-compute a bulk of the averaging needed to map the texture onto the smaller space. We can immediately select the mipmap texture that matches the size of the desired pixel thus allowing us to decrease the amount of averaging work we need to do to assign the pixel a color.

jasonTelanoff

Is there ever a case when a mipmap of more pixels than Level 0 would be constructed. It makes intuitive sense that it wouldn't, however if you want to render something with more pixels you'd have to use a previous technique for each one, rather than just pulling from a stored image.

litony396

Is there ever a point in which a mipmpap level that goes to 1x1 isn't enough to be effective at antialiasing? In this case would a higher resolution Level 0 image be able to fix that?

antony-zhao

Similar to @jinweiwong I've seen mipmaps mentioned in video game settings so it's super cool to see the exact intuition behind it. Also @brandonlouie's explanation helps make sense of it as well.

Alescontrela

This is an extremely clever idea that at first may seem too expensive given the additional memory necessary to store all the images, but once you optimize the storage of the mipmap it becomes quite efficient.

You must be enrolled in the course to comment