No announcement yet.

Ft Pattern Suppressor

  • Filter
  • Time
  • Show
Clear All
new posts

  • Chain and Maple,

    Internal working of the Fourier routines is floating point or 32 bits/color/pixel. Also the code is limited to 32767 px on the longest side - worked on increasing to 300000px but Adobe wouldn't share documents about what to do so stopped trying. The Pattern Suppression also pads the image by either 10% or 100px on each side (Chain which is right? My memory is not very good anymore.) .

    I'd suggest windowing the file with a large overlap. You have 27000x54000 px image, Add a 25% inner overlap of the windows. I'd set windows at x 0 to 27000/2+3400 and 27000/2-3400 to 27000 and y do likewise but maybe do 3 windows so the windows aren't too large. Run the suppression on the 4 or 6 windows and then piece them together with PS.

    The 300dpi doesn't enter into this. You should gauge that by the viewing distance.

    Let me know if you need more help.


    • ​​Meticulous painting,Water-based ink on silk ,in Chinese -->(绢本水彩工笔)

      in the past, my workflow was exactly the same with you suggested(@Chain @Ron Chambers):

      Capture with a camera, filtered separately,then combine ,

      Just worried about regional imbalance

      PS: Affinity Photo -> denoise -> fft denoise, can do the process,but cannot save back to PSB




      [processed screenshot]

      Anyway, thank you again for your great job !

      If you have the opportunity to come to China, look forward to meeting in Beijing. ^_^


      • I would suggest that you always filter before combining. When you combine first the "noise" must be consistent which it most likely won't be. If the noise is not periodic from panel to panel you get a poorer suppression. Keep in mind that the Fourier Transform is global in nature but each of your shots are independent. Size isn't as much a problem with the individual images. You should overlap the camera shots also.



        • I see! And you are trying to get rid of the silk fabric pattern. Quite insanely large images.

          I have to agree with Ron here that it will most likely be best to run the suppression before combining the images (due to the pattern as well as the image size).

          If you are somehow getting "regional imbalance" from one image to the next so they don't match, it would be possible to reuse the suppression layer from one image on another if you want to perform the exact same suppression (but this probably only makes sense if the pattern to be suppressed turns out identical in both after the fourier transform). You could also consider copying/reusing only the mask (perhaps a better idea).


          • Position of the pattern in each image dictates that you do the suppression separately. Are you doing lense corrections etc? What about having the camera plane parallel to the fabric? The slightest tilt or twist in either the camera or the fabric changes what the transform sees. Even though what is being applied is an amplitude effect, the computation of the transform has to have equivalent phase or positioning. What does "regional imbalance" look like? You may be suppressing information too near to the zero frequency position in the transform. I'd recommend that you save off the the transform before picking and after picking for the individual images and view them closely and side by side.

            Do you really believe that the fabric is stretched the same over the entire thing? It is probably different within each image. You may need to only use the sweet spot of the lense which means you have to take more shots. How consistent is the lighting across the image? Any shadows due to the building structure or lighting? Is there curvature in the wall? Temperature and or humidity changes will change the stretch of the fabric.

            I've set up shooting of fine art and the problems are never ending and due to the nature of the art you have no control.

            Do you have pixel shift available in the camera? It or something equivalent is needed because of artifacts from the suppression of the Bayer pattern add low and high spatial frequency noise. Remember here that you are pixel peeping with the whole experiment. You have to handle all of the little things that are ignored in most photography.

            Last edited by Ron Chambers; 09-19-2020, 05:57 PM.


            Related Topics


            • visnew
              Virus / Spyware in plugins or actions
              by visnew

              I'm not sure if I'm posting in the right place, but since the previous "Virus" thread was posted here, I thought it was best not to change.

              In the original thread, Gary warned us not to download stuff from unknown or dangerous sites. However, I'm wondering...
              05-15-2008, 02:42 PM
            • Rydiant
              Is it worthwhile to sell actions?
              by Rydiant
              I've developed a series of actions that produce effects like some commercial plugins. Because the quality of output compares well (and in some ways surpasses what's already out there,) I think users would be willing to pay for the actions at a fraction of the cost of a plugin alternative.
              04-18-2010, 05:24 PM
            • GaryGornelli
              by GaryGornelli
              Hi, I was discussing brushing up work to show clients the results at shoots and was told that some photographers run a few actions. Curves, smoothing/sharpening etc. Do any of you? What are they?

              07-16-2009, 11:28 AM
            • Sanda
              strange behavior
              by Sanda
              After a recent hard drive crash and replacement I had to reinstall Photoshop 6. I've reinstalled all of my plugins and actions ect. All seemed to be working fine but today I needed to use an action for the first time since the reinstall and to my amazement discovered that my actions pallet won't...
              09-18-2002, 03:26 PM
            • Doug Nelson
              Creating Actions in Photoshop
              by Doug Nelson
              View here
              01-11-2003, 07:34 PM