[Python] How to evaluate 2 images with different size due to downscaling task

Discussão em 'Python' iniciado por Stack, Março 17, 2025.

  1. Stack

    Stack Membro Participativo

    I have a netcdf4 file contains the original satellite data. I performed 3 different types of algorithms to downscale it (convolution, gaussian with bicubic, area averaging). Now I want to find a way to evaluate how "well" the algorithms perform. Note that the original and the downscale have different size (1472x1069x1069 vs 1472x528x528).

    Is there a way to do this? I have read through SSIM, PSNR, even for MS-SSIM, VIF and LPIPS, I notice that they all require similarity in shapes. I also don't want to resize my images to preserve the quality.

    Thank you!

    Continue reading...

Compartilhe esta Página