AHD vs. VNG RAW interpolation

[tomczak]

I played a bit with the AHD and VNG interpolations for the same RAW files and I don't find them that overly different. With low noise images, AHD seems to produce less zipper-like artifacts on sharp edges than VNG, but with high noise images it produces these unnatural mosaic-like structures in smooth areas (VNG makes noise look more natural, but also produces some single-pixel 'spikes' in the noise).

What are other's experiences with it?

[mjdl]

The forums at the RawTherapee raw processor software website http://rawtherapee.com/forum/ have dozens of threads comparing many kinds of raw demosaicing algorithms. Maybe a bit of a time-waster to find good discussions among all the Extreme Pixel-Peeping sportsmen, though.

There's always Guillermo Luijk's in depth tutorial for DCRAW at http://www.guillermoluijk.com/tutorial/ ... dex_en.htm which goes into depth on DCRAW's raw image process techniques and options. PWP50 seems to expose some of the options discussed there, so it's useful for background.

[den]

For me in general, AHD interpolation seems to provide the best, all around results. The only time I resort to VNG is when I have exceeded a len's diffraction limits during a 'capture'. In this case, I will do two RAW conversions, one with AHD and the other with VNG... ...and use the conversion that will best provide a final preference image when recovering using various sharpening techniques from the softening due to the lens diffraction... ...in some rare cases, I have specific image area Composite-blended the final preference images from the AHD and VNG conversions.

[tomczak]

Den, what's the visual/practical difference between AHD and VNG interpolation when developing diffraction-blurred RAW? Why would you bother to use VNG, only for blurred RAW? Is it because of aggressive sharpening and the 'maze-like' AHD noise being emphasized and with sharper initial RAW you wouldn't have to go as hard on the sharpening sliders so the 'maze' doesn't show that much?

OTF: how do you attempt to sharpen such images?

[den]

MaciejT....

(1) Never really did consider the differences. AHD and VNG interpolation conversion images just provided differing starting points for specific image area sharpening [and occasionally specific image area noise reduction] and other post-processing techniques. Did not ascertain generic relationships that could significantly be due to the method of interpolation. My overall impression was that the subsequent post-processing techniques obscured the differing interpolation artifacts so that they became essentially a non-factor.

It has been some time since I did this and I was unable to locate my notes... in any event, my current photographic style [Canon EOS-350D] is too limit aperatures to f11 and lower with the camera on aperature priority [usually f8 which seems to be a good 'camera ready initial' setting for compromise DOF and in-focus 'sharpness' considerations] and if I need more speed, to up the ISO from 100... ...so it has been some time since I really had to deal with lens diffraction issues.

(2) lens diffraction recovery sharpening and other difficult recoveries... In addition to the usual suspects: USM, BilateralSharpening, etc. and depending on the desired output resolution: (a) downward Resizing for web image resolutions will often recover from lens diffractions after a 50% to 65% Amount of Sharpen or just printing on 4x6 inch paper letting the printer driver do the sharpening; or (b) for gallery quality prints and 3456x2304 pixel image resolutions, I try:
(i) micro-contrast sharpening with AdvancedSharpen with a setting [*.as file in txt format] and sometimes with the suggested mid-tone mask described in (ii):

AdvancedSharpen 1.0
speckradius 0
speckthresh 255
speckdiffthresh 255
removewhitespecks 0
removeblackspecks 0
blurradius 1.0000
bluramt 0.0000
blurthresh 255 255
blurflags 0
sharpenradius 0.4990
sharpenamt 1.2012
sharpenthresh 4 8
sharpenflags 1
histexpand 2
end

adjusting settings as needed and followed by if necessary, black/white speck removal of 1 pixel of the resulting image version using AdvancedSharpen's Speck Removal tab;

(ii) then HighPass Filter-Hard Light sharpening with a 3 to 5 pixel radius; followed by HighPass Filter-Hard Light sharpening with a 50 to 75 pixel radius [Local Contrast Enhancement]. Both of the HighPass-HardLight sharpenings use masks to limit the effect to mid-tones and to limit the white half of the sharpening to 2/3's of the black half of the sharpening.

The mid-tone mask [Composite-HardLight Input mask] is a Brokenline curve: [0,0], [5,0], [20,40], [50,100], [80,40], [95,0], [100,0] and the mask to control white/black half sharpening ratio [Composite-HardLight Overlay mask] is the default diagonal [0,0], [100,100] with the white amount set to 66.7% and black amount set to 100% when used in the Composite-Hard Light transform.

These two masks when used, will allow the use of the Composite-Hard Light filter rather than the normal Soft Light filter when employing High Pass Filter sharpening sequences without generating objectionable white sharpening halos...