Description of the desired feature
Similar to Bekaert et al 2020, I would like to implement masking of velocity data based on velocityStd data according to the formula: |velocity| > (2*velocityStd) . This masking procedure is helpful when trying to isolate areas that have a high magnitude of local deformation.

Describe the solution you'd like

I can think of two approaches for how this could be done within generate_mask.py, but I would love to know which method @yunjunz thinks is best!

1. Implement the entire masking procedure at once, which would not allow a high degree of user modification and restrict this masking process to velocity.h5 datasets
2. Implement each portion separately (see bullets below) so that future mask generation features can use some of the functionality
   • Add absolute value option to image_math.py
   • Multiply velocityStd by two using existing functionality
   • Allow the -m and -M options of generate_mask.py to accept filenames with a dataset specification

Are you willing to help implement and maintain this feature? Yes/No
Yes, I am willing to implement this improvement as soon as we determine which method we would like to use!