MinMax, Percentile, MinMax+FT, and Percentile+FT implementation?

[ming053l]

Hi, thank you for sharing this great work! I have some questions regarding Table 3 in your paper:

How were the MinMax, Percentile, MinMax+FT, and Percentile+FT methods implemented in your experiments?
For "+FT" methods, how did the fine-tuning process work? Did you use the calibration dataset to optimize the parameters?
Additionally, I noticed that some models are marked with "FQ" (Fully Quantized) while others aren’t. Could you share how being fully quantized influences the PSNR/SSIM and computational efficiency?

Thanks in advance for your time and for sharing this research. I’m looking forward to your reply! 😊