ACR Super Resolution

Discussion in 'Digital Darkroom' started by tonybeach_1961, Mar 19, 2021.

  1. [​IMG]

    A 100% crop from the D800 on left, and to its right D500 resized to 36 MP, and in the center a D300 file resized to 36 MP. Second from the left is the native D500 file and to its right the D300 file resized to 21 MP. I also used Topaz NR set to Low for noise reduction and for enhancing. along with Smart Sharpening on all the files (then gave some of the sharpening back when I compressed the crops to a manageable file size for displaying here).
     
  2. I don't understand this. ACR super resolution doesn't offer a choice of resizing amounts. It automatically generates an image with 2 x the number of pixels in both directions. So shouldn't the D300 image, for example, be 49 MPX after super resolution? I must be missing something.
     
  3. It shows that nothing beats real resolution.
     
  4. digitaldog

    digitaldog Andrew Rodney

    Correct. The result is this resized image, using AI, as a Linear DNG.
     
  5. What your missing was my omission to explicitly state my approach, so my apologies for that. I wanted to arrive at equal output, so after the algorithm did just what you described I downsized the files to match the various other outputs. My assessment is that the files look rather bad fully upscaled, and besides that my goal is to prepare them for a printer at 300 ppi.
     
  6. I have read that while upscaling, some prefer to avoid generating new colors and use simple Imagemagick scale operation to resize image. This sometimes leads to pixellated output, but there is no grayish messiness.
     
  7. I'm thinking I should further clarify what the crops show. The three on the left are all 36 MP and the two on the right are 21 MP. The numbers displayed at the top are the original MP (I should have just written, D800, D500, and D300).
     

Share This Page

1111