Using Pearson correlation coefficient as a shape descriptor

https://en.wikipedia.org/wiki/Pearson_correlation_coefficient

Hello, as you can read in the page reported above, the pearson correlation coefficient could be used as shape descrtiptor. I would like to use that method to extract some features from a segmented image: in particular, I would like to distinguish element with an approximate linear shape (r=+-1) from the other elements (r<<1). I have an image with several segmented objects and I would like to evaluate that coefficient per each object.

Please can you help me?