I am working on a Python script for dark object subtraction, a radiometric correction method for remotely sensed data.To accomplish this, I need to find the minimum band pixel value of a waterbody within a region of interest (ROI) (screenshot).  My initial thought was to create a polygon within a Landsat scene and use Zonal Statistics as Table to find the minimum value.  However, I cannot easily call this minimum value later on in the script.  Another thought was to convert the band to a numpy array and find the min value (excluding 0), although I am uncertain how to subset the array to my ROI.  How can I find the minimum value using a region of interest so that I can easily call this value later on in the script? 

![enter image description here][1]


  [1]: https://i.sstatic.net/4Nry8.jpg