Many years ago, a talented and iconoclastic engineer named Bob Carver issued a challenge to the staff of a “golden ear” hi-fi magazine. They could pick any amplifier, and he would attempt to make his distinctly mid-fi $700 amp sound so much like it that the staff couldn’t tell them apart. They hemmed and hawed, and finally accepted. Carver won.
You can read about it here: http://www.stereophile.com/content/carver-challenge
That got me thinking. Not about making an optical silk purse out of a sow’s ear, but about using software to take images from very good lenses with few departures from perfection – attractive or otherwise — and make them look like they were taken with lenses that leave distinctive fingerprints on images made with them.
It’s not an entirely original thought. You can buy iPhone apps and Photoshop plugins that simulate a Holga camera image, lens distortion, flare, light leaks and all. But has anyone done it for lenses that photographers cherish for their character? And are there lenses good enough to serve to make the original image? Does anyone want to write the program to do the modification?
And, if done, does anyone want to try and tell which image was made by the real famous characterful lens and which was massaged into loveliness with software?
I don’t know if this is going anywhere, but it’s fun to think about. Does someone want to nominate a legendary full-of-character lens to be our target? We need to be able to find one copy of whatever it turns out to be.
The challenge with lenses is that they respond to a 3-dimensional scene, and no matter how perfect the starting lens is, you need the 3d data (you must at least focus stack the entire DOF from the nearest object all the way out to infinity) to fully reproduce the “character-y” lens. This is especially true of longitudinal chromatic aberration.
Good point. But maybe we can get close enough for some kinds of lenses.
Andre Y says
Jim, I was reading the LensRental test of the new Canon 11-24 lens (http://www.lensrentals.com/blog/2015/02/canon-11-24-f4-l-mtf-tests), and was impressed by how tightly grouped its MTF curves were, which are basically a rough approximation of the lens’s frequency response.
In the case of the Canon, it has relatively flat frequency response (or at least less downward-trending than the other lenses it was compared again), and that should affect how we perceive a subject with details of varying sizes, I would guess.
I wonder how much of lens character is determined by its frequency response, much in the same way that an audio component’s sound is determined by its frequency response.
Erik Kaffehr says
Once we have good enough resolution to avoid aliasing artefacts it should be possible, at least in theory, to deconvolve an image, if the PSF is known or well approximated. Another PSF could than be applied to the deconvolved image.
I think this is entirely possible for planar objects, but not possible with 3D objects and out of focus rendition, as different PSF would be needed at different depths.
Neslon Pass, an amplifier designer of the very highest level, has some interesting thoughts on distortion.
I am a Nelson Pass fan. I have one of his Aleph class A amps, and used to have his Stasis ones, and the first Threshold preamp, with internal factory mods to FET-1 circuitry.
Somehow I knew you would be a fan.
The Aleph is a durable milestone in design that is rare in any field.
His talks at Burning Amp are on youtube and very interesting.