No announcement yet.

Deconvolution Sharpening

  • Filter
  • Time
  • Show
Clear All
new posts

  • Deconvolution Sharpening

    Click image for larger version

Name:	image_3380.jpg
Views:	0
Size:	604.9 KB
ID:	9071

    ​ ​The image on the left is the original .NEF (raw) image from the Nikon D800 converted using Adobe Lightroom and converted to a .jpg image. No other processing was done and all default options were switched off. The image on the right is the same image with the exception that deconvolution sharpening was applied using the ;Details' panel before the image was converted to a .jpg image. The Sharpening settings in the detail panel are:
    Amount : 86
    Radius: 0.5 pixels
    Details: 100
    Masking: 0

    With the Details slider set to 100 (extreme right side) deconvolution sharpening is applied. If the slider is to the extreme left the Unsharp Mask sharpening is applied - anywhere in between is a blend of the two sharpening methods.The major difference between the 2 sharpening methods is that USM applies a gaussian blur during the sharpening process whilst deconvolution sharpening applies a lens blur. I have found very little in the way of reliable information as to explain exactly what effect the different blurring is controlling.

    Comparing the 2 images above, I can concern 3 major differences:
    1. The 'sharpness' in areas of fine detail is greatly improved. This clearly evident in the foliage of the trees on the left
    2.Overall the colours are more accurate. The bluish colour cast caused by the dense shade in the foreground and the atmospheric haze in the background has been largely eliminated
    3. The tonal range appears to be more balanced with more natural transitions between shadows, midtones and highlights

    The next 2 images are included to give you an indication of how this process transfers to the finished image. The first is a 'completed' image and the second is a cropped at 100% zoom to show the level of detail in the hills in the background which at a guess would be 10+ miles away.

    Click image for larger version

Name:	image_3382.jpg
Views:	1
Size:	464.9 KB
ID:	9072

    Click image for larger version

Name:	image_3381.jpg
Views:	0
Size:	538.1 KB
ID:	9073

    I would appreciate your comments and comparisons with any experiments that you have undertaken

  • #2
    Somehow we managed to lose the last paragraph in the above post -

    One other thing that I should mention is that things went pear shaped very quickly if I was too heavy handed with the amount of sharpening that I applied. In reality, an amount of about 50 would be more appropriate, but when the image is down scaled from 36mega pixels to the maximum size permitted by the forum, you would hardly detect the changes.


    • #3
      Round Island at the Nobbies, Phillip Island, Vic - taken on the Australia Day weekend

      Click image for larger version

Name:	image_3390.jpg
Views:	0
Size:	406.3 KB
ID:	9082

      Composite image using 3 images taken over a period of about 1 minute to capture the continual movement of the ocean crashing onto the rocks from the foreground to the background. Images were manually blended together and the 3 step sharpening technique mentioned in post 22 of the "FORUM SHUTDOWN yes or no" thread incorporated into the workflow.

      Camera Nikon D800 with Nikon 20mm f2.8 lens f13 @1/160sec. Processed in Lightroom Classic CC and Photoshop CC 2019. Capture (deconvolution) sharpening was applied in Lightroom before the images were blended (using a combination of painted layer masks and Luminosity masks) and local adjustments made in Photoshop. Final sharpening was High Pass filter. Image size before resizing and conversion to sRGB for posting on the forum was 7377 x 4946 pixels.


      Last edited by oldgreybeard; 4th February 2019, 12:12 PM.


      • #4
        I didn't realise how much my cataracts effected my image processing.

        Looking through some images which I had processed before my cataracts were removed, I was surprised at how much of a yellow /red cast I had introduced into the image.

        The first image was processed before the operations and the second today.

        Click image for larger version  Name:	Clydesdale-1.jpg Views:	3 Size:	89.4 KB ID:	9093

        Click image for larger version

Name:	Clydesdale-2a.jpg
Views:	0
Size:	105.2 KB
ID:	9096

        I guess the logical question is how can I be sure that today's image is correct? The simple answer is that I have learnt to rely much more on the numbers rather than my eyes. The "info" panel is my best friend.

        The following screen shot was taken when I had finished processing the image (removing unwanted elements and blurring the background to increase the viewers attention on the horse.
        Click image for larger version  Name:	Clydesdale-3.jpg Views:	1 Size:	203.6 KB ID:	9095

        In order to correct the colours, I needed to find a point in the image which I knew to be a neutral colour. The point identified by the colur sampler on the nose band of the bridle is such a point. It is a dark neutral grey and such should have equal values of Red, Green and Blue. Using a curves adjustment layer, the individual (RGB) channels where adjusted. The initial reading from the info panel showed the red channel was 85, the green 77 and the blue 78. The red and blue curves were adjusted until they each read 77, matching the green channel.

        The image is now correctly colour balance based on the dark neutral gray of the bridle. This is a very effective way of determining the correct colour balance and removes the 'guessing' of trying to do it by eye - particularly if you have vision concerns.

        The image was cleaned up using a Wacom Tablet to clone out the old caravan a in the background and the reins which were hanging down in the foreground.
        Attached Files
        Last edited by oldgreybeard; 8th February 2019, 07:52 PM.


        Unconfigured Ad Widget


        Unconfigured Ad Widget