Fig. 9
From: Orthogonality is superiority in piecewise-polynomial signal segmentation and denoising
Example of time-domain reconstruction, test signal “1”. Left side shows the noiseless and noisy signals, the plot on the right hand presents the direct signal estimate \(\hat {\mathbf {y}}=\mathbf {P}\mathbf {W}\hat {\mathbf {x}}\) (CA), and the respective least squares refit (LS), on top of the noiseless signal. Clearly, LS radically improves the adherence to the data (and thus improves the MSE). The bias of the CA is explained in Section 3.2