# Can coefficients of a DWT be computed using least squares approximation?

I assume that it can be computed given function values for each sample as a vector and wavelet values for each scale and translation packed in a matrix.

Can anyone clarify the matrix construction and if its even possible?

• I must admit I don't understand – the DWT is a linear operation, and thus, the least squares approximation of it is the DWT operation itself (leading to an estimation with zero variance). The DWT is a linear operation on vector spaces – it hence can be represented as matrix. – Marcus Müller Mar 15 at 8:28