ophys_etl may have additional metrics that could be useful
Notes 7/19:
Suite2p uses a PCA-based approach to find the most varying frames in the registered movie. It compares them to see whether the variation is due to neural activity or uncorrected motion. For each of the 30 top PCs it averages the top 300 frames and the bottom 300 frames and registers the resulting 2 images. rigid, average nonrigid, and max nonrigid shifts are saved as "regDX". Those scores should all be very close to zero.
Save those PCs "tPC", "regPC" as well as "regDX".
Use the same PCA-based approach, but use Farneback method for registration instead of Suite2p. Save the residual optical flow "farnebackROF" and "farnebackDX".
Also store the "crispness", i.e. the norm of the gradient (acutance) of the mean image for the original and registered movie.
For non-rigid registration create a summary image (showing shifts and phase correlation) for each patch.
Notes 7/12:
Notes 7/19: Suite2p uses a PCA-based approach to find the most varying frames in the registered movie. It compares them to see whether the variation is due to neural activity or uncorrected motion. For each of the 30 top PCs it averages the top 300 frames and the bottom 300 frames and registers the resulting 2 images. rigid, average nonrigid, and max nonrigid shifts are saved as "regDX". Those scores should all be very close to zero.