I use the Transport of Intensity Equation (TIE) algorithm to retrieve the phase of an image. To evaluate the performance of this method, I use the RMSE (Root Mean Square Error). However, the RMSE values do not seem to accurately represent the algorithm's true accuracy.
Currently, I calculate the RMSE using the following methods:
- Theoretical RMSE: [ \text{theoretical_rmse} = \sqrt{\text{error_mean}^2 + \text{error_variance}} ]
- General RMSE formula: [ \text{rmse} = \sqrt{\text{mean}((\Phi_{\text{DCT}} - \text{Phase})^2)} ]
Is it possible to normalize the RMSE by dividing it by the maximum pixel size and (\pi) to obtain a more accurate result?
Thank you!
