What problem does your feature request solve?

A fundamental aspect of evaluation between two models, and indeed verification, is how similar two objects are. For meteorological purposes we are often comparing two charts, and more qualitatively are doing "spot the difference". Image processing techniques could be beneficial in quantifying these differences, and make a first move in quantifying "fragmentation" of objects. The structural similarity (SSIM) index is one such diagnostic that could be used. It approximates human perception (albeit not fully) by considering three aspects: luminosity, contrast, and structure. It combines these to create a score to indicate how visually similar two images are. A single number or a 2D map of the fields can be created.

Describe the solution you'd like

The SSIM is available through the sci-kit image python module, and thus only minor additions are needed to convert it into an operator for CSET: namely creating a cube output as part of a function that calls the sci-kit image module.

Further advantages of using sci-kit image will be to enhance cell statistics code #329 and allow other statistics to be produced.

Describe alternatives you've considered

A wealth of image processing diagnostics exist, and this is one of many that could be considered.