I have a situation where I am sent a series of photographs of hand-written documents (quite old, from 1850-s). Each photo is done just as you and I would photo a page from a book, so not a professional environment. So I have 3-4 images per page, all with a slightly different .. angle, lightning, etc. But the most tedious part that I hoped to automate was "to find which of the images in the set is best
focused to be able to actually read the text"
So I've did some search here and on the Google, and people were suggesting to use such metrics as "deviation" and "enthropy". Now, I use
imagemagick to get those values, and it seems as they are absolutely irrelevant for the task at hand, i.e. the images I would personally choose as best quality, have quite average values, for any given set of images & for both metrics (deviation/entropy), i.e. no min, no max, nothing. I am lost.
Maybe there's a (hint of) the solution, but I just can't google it properly ? Thanks for any suggestions