Would the weights on the laplacian stack be computed based on the amount of details or gradient in the layer of the laplacian images? If so, are the weights for the higher frequency layers of the stack sensitive to noise?

I think the weights in weight map for Laplacian pyramid can be blurred based on the frequency band of the level. High frequency band needs less blurring of weights at the edges to preserve the hard edge details in the final image.

Are these weights in weight map computed with the heuristic like having greater weights in bright region of an underexposed image? If so, would the blurring of weights use Gaussian filter to adjust the amount of blurring based on the level of weight map?

In the lecture, Prof Kayvon mentioned that this approach blurs different layers of laplacian pyramids by a different amount. Could we interpret it as that it blurs the weight map on low-frequency content more than that on high-frequency ones?