Multiple sharpening, frequency and order

[tomczak]

Den uses an interesting multiple sharpening with different radii in his example of supervised workflow:

http://www.ncplus.net/~birchbay/tutoria ... /index.htm

There is a few things I wonder about:

- each USM causes some halo of different frequencies - don't they add up visually?
- while changing the radius focuses the USM effects on different frequency features, it sharpens the other frequencies too, just not so much. What would be the best sequence then of such sequential sharpening. For instance would it make more sense to sharpen Large-Medium-Small, or as Dan suggests: Medium-Small-Large, or some other configuration?
- why BilateralSharpen at the end? - isn't it almost the same as USM1 (medium radius), except that it protects some major edges from halos (but the USM that doesn't protect them has been already executed a few steps back).

Cheers!

p.s. Also, why using the RGB mode in Brightness Curve would be preferred over HSV-V?

[den]

There is a few things I wonder about:

- each USM causes some halo of different frequencies - don't they add up visually?

Complete elimination of the light half of USM is not necessarily a good thing or a preference. Note that the USMSharp(3) sequences have: (1) Low Amounts: 40, 50, and 17.5; and (2) the Radius = 6 and 1 USMs are used with a Threshold = 3 which tends to soften artifact type light/dark sharpening edges. As for adding up... one can soften their effect with the down stream BilateralSharpen widget if a preference.

- while changing the radius focuses the USM effects on different frequency features, it sharpens the other frequencies too, just not so much. What would be the best sequence then of such sequential sharpening. For instance would it make more sense to sharpen Large-Medium-Small, or as Dan suggests: Medium-Small-Large, or some other configuration?

??? ...a Medium-Small-Large sequence with the described USM settings/Amounts previously described generally result in obtaining preference details and contrasts for my images without objectionable/noticeable halo or artifacts...!!!

- why BilateralSharpen at the end? - isn't it almost the same as USM1 (medium radius), except that it protects some major edges from halos (but the USM that doesn't protect them has been already executed a few steps back).

Chromatic Abberation correction tends to soften image detail. The BilateralSharpen widget at the suggested Amount/settings can re-capture detail being careful to not let it produce objectional light halos.

Cheers!

p.s. Also, why using the RGB mode in Brightness Curve would be preferred over HSV-V?

Image data is recorded in the RGB color space model, it is converted/gamma expanded in the RGB color space model... ...so why not continue the process when further aggressive 'luminance' changes are needed including saturations that change as well in accordance with the RGB color space model dependencies. The same aggressive changes made to HSV-V without an accompanying HSV-S change generally results in HSV-S artifacts where saturation contrasts have no blending gradients. HSV or HSL edits seem to work best when 'fine tuning' a RGB converted/gamma expanded and aggressively changed resulting image...

In any case, the validity of the Workflow is not so much the theory applied to constructed test images but to real-life actual recorded image data... Have preferences been achieved?... Details preserved?... ...You be the judge...

Below are 1:1 resolution, 200x300 pixel image area crops, unchanged accept using jpeg compression so that the file size is 50KB or less, of the resulting Workflow image. The starting file is a Canon 350D CR2 file, 3456x2304 pixels, recorded with 'good' photography technique: tripod; f11 to reduce lens defraction; exposure checks to ensure the recorded data is within the sensor's range; ISO 100 for lowest noise; exposure not right shifted to retain low contrast cloud/snow detail...

wfl_1-1area_samples_1.jpg (44.15 KiB) Viewed 1561 times

wfl_1-1area_samples_2.jpg (47.59 KiB) Viewed 1560 times

Additional edits depending upon Compositional Crop locations that the resulting Workflow image might need, would be upper left and right corner light falloff and lens distortion or warp correction of the inward leaning left and right side trees.