An empirical assessment of the structural similarity index

This research evaluates the performance of the structural similarity index. An increasingly popular image quality metric, the structural similarity index is essentially a measure of the statistical correlation of two images. Its alleged strengths and advantages are widely acclaimed. For an arbitrary set of test images, however, statistical regression reveals a surprisingly high degree of association between the index and the fallible mean squared error, despite claims to the contrary on the part of its designers. This discovery suggests a much closer relationship between these two metrics.