vPSNR calculation

Visualization-aware image fidelity measurements

Image fidelity metrics are essential in order to evaluate the effects of data reduction methods. Traditional fidelity metrics have drawbacks that make them inappropriate for many actual clinical situations. More specifically, the following three requirements are necessary for general purpose clinical image fidelity metrics:
- High-dimensional: The metric must not be restricted to 2D images
- Visualization-aware: The metric must account for visualization transformations
- Reduced-reference: The metric must be independent from the original data

The metric vPSNR (which stands for "Visual PSNR") is a PSNR variant that
accounts for visualization effects in an explicit and well-defined way: the sample-based differences are assessed according to a) the value transformed by the visualization and b) a perceptually correct scale (CIELAB). We provide code for both exact and approximate calculation, where the latter is a reduced-reference approach based on simplified histograms.

Please refer to the following paper for full information about the method:
"vPSNR: A Visualization-Aware Image Fidelity Metric Tailored for Diagnostic Imaging", International Journal
of Computer-Assisted Radiology and Surgery.