Image selection for PSNR and SSIM indicator calculations

[1085437068]

Thank you very much for your contribution. I have a question about the calculation of PSNR and SSIM in the paper. Calculating these two indicators requires two pictures. In addition to selecting the processed picture, which picture do you select as the other picture?

[1085437068]

In fact, when I calculated the PSNR using the generated image and the content image, it was only about 18 dB. Could the author explain the calculation process for this metric?

[siddique-mab]

Hey @1085437068, You should compare the original image with the generated image. As far as I remember, depth contribution could greatly vary these metrics values. When working with Ustyle, please play with depth contribution to get the desired outcomes. Also, these metric values could vary greatly with a particular underwater image quality. A few severely degraded underwater images could greatly pull down these values. In all our comparisons with all the baselines, we removed these outlier images if there were any. Please let me know if you have more questions!