eduzhai > Applied Sciences > Engineering >

Understanding SSIM

  • Save

... pages left unread,continue reading

Document pages: 8 pages

Abstract: The use of the structural similarity index (SSIM) is widespread. For almosttwo decades, it has played a major role in image quality assessment in manydifferent research disciplines. Clearly, its merits are indisputable in theresearch community. However, little deep scrutiny of this index has beenperformed. Contrary to popular belief, there are some interesting properties ofSSIM that merit such scrutiny. In this paper, we analyze the mathematicalfactors of SSIM and show that it can generate results, in both synthetic andrealistic use cases, that are unexpected, sometimes undefined, andnonintuitive. As a consequence, assessing image quality based on SSIM can leadto incorrect conclusions and using SSIM as a loss function for deep learningcan guide neural network training in the wrong direction.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×