A Formal Assessment of the Structural Similarity Index

In recent years the structural similarity index has become a de facto standard among image quality metrics. Made up of three components, this technique assesses the visual impact of changes in luminance, contrast and structure in an image. Present applications of the index include image enhancement, video compression, video quality monitoring and signal encoding. As its status continues to rise, however, so do questions about its performance. In this paper, it is shown, both empirically and analytically, that the index is directly related to the conventional, and fallible, mean squared error. This relationship raises serious questions about the credibility of the structural metric. Index Terms structural similarity index, SSIM, mean squared error, MSE, image quality metric.