I'm not exactly sure why the first behavior is not what we expect - if we have many translucent things overlapped on one another, shouldn't we expect the total to be darker like in the first case as each of them would block some light? At least physically this would make more sense

kkzhang

Does this mean that it doesn't matter what order two triangles are composited over one another in the premultiplied case if the two triangles have the same exact color?

frogger

I have a similar confusion to one of the above commenters: shouldnâ€™t it be the case that having multiple overlapping translucent images is different than having just one? It at least seems to be the case in the image shown on this slide: for example, if we look at triangles A and B, the region where they overlap is darker than either individual triangle.

air54321

Does this sort of calculation generalize to a lot of different objects? Do we essentially just directly work with all objects at the same time, or is there ever an instance when we want to break down many overlapping images and consider each pair of them separately?

bobzhangyc

For the non-premultiplied, would it be brighter if we take the average of the two alphas?

David

In the non-premultiplied case, do we use alpha only for multiplying and not for rendering? If we multiply alpha with color when we render like the premultiplied, wouldn't the color be lighter than we want it to be?

I'm not exactly sure why the first behavior is not what we expect - if we have many translucent things overlapped on one another, shouldn't we expect the total to be darker like in the first case as each of them would block some light? At least physically this would make more sense

Does this mean that it doesn't matter what order two triangles are composited over one another in the premultiplied case if the two triangles have the same exact color?

I have a similar confusion to one of the above commenters: shouldnâ€™t it be the case that having multiple overlapping translucent images is different than having just one? It at least seems to be the case in the image shown on this slide: for example, if we look at triangles A and B, the region where they overlap is darker than either individual triangle.

Does this sort of calculation generalize to a lot of different objects? Do we essentially just directly work with all objects at the same time, or is there ever an instance when we want to break down many overlapping images and consider each pair of them separately?

For the non-premultiplied, would it be brighter if we take the average of the two alphas?

In the non-premultiplied case, do we use alpha only for multiplying and not for rendering? If we multiply alpha with color when we render like the premultiplied, wouldn't the color be lighter than we want it to be?