If you are using bilinear interpolation to upsample but Gaussian blur to downsample, won't the reconstructed image using the Laplacian pyramid be slightly different?

we can see from L4, L3, L2, L1, L0 are gray, is it because the calculation from L4 = G4-up(G5) lost some information?

what's the up() expression here suppose be?

How far does the pyramid go? Is there a limit or does it generally stop on L5?

I think up is upsample

Can reconstruct original image by upsampling L5 to L4's size, then temp = L5 + L4. Then upsample temp to L3's size, temp += L3 etc.. until temp = temp += L0 == original image

@lwzt this is probably just done for convenience/cost reasons - probably hard to "upsample" by "inverting" the gaussian (and empirically unnecessary)