For the TC paper, we used residual correlation of alignment points as a quality metric for the alignment: Low correlation of residuals means high independence of residuals and therefore good image alignment. We could publish this number in the statistics file or in the meta file for alignment. An improvement could be made by performing a normality test on the residuals or look at the variance-covariance matrix after the last LSM iteration. This would actually test for the required assumption (for LSM) that residuals are i.i.d. normally distributed.
For the TC paper, we used residual correlation of alignment points as a quality metric for the alignment: Low correlation of residuals means high independence of residuals and therefore good image alignment. We could publish this number in the statistics file or in the meta file for alignment. An improvement could be made by performing a normality test on the residuals or look at the variance-covariance matrix after the last LSM iteration. This would actually test for the required assumption (for LSM) that residuals are i.i.d. normally distributed.