Are higher quality resizing algorithms worth it?

[xStrom]

Need to figure out if higher quality resizing algorithms are worth it over just using bilinear.

What's the performance like, would a flash of a lower quality resized image into a lazily loaded higher quality be an acceptable visual experience?