Brick wall test

Did the brick wall pass or fail?

Or is further testing required?

No one wants to play GB 😛

If that is a test of somebody's bricklaying skills, I would say that he passed with a good mark.

Don

Mortar aliasing present, but reasonable.

Good. I would try something like Zig-Align and dial back sharpening a little.

Superb micro-contrast - the mortar only 1 px wide !!

GB? Is that a game?

Very close to 1px The mortar lines vary thickness in a periodic manner, more obvious at 400% than 100% on my screen. Any less than 1 and the under-sampling would show a lot more artifacts, such as the mortar being too dark and brownish, periodically. At 100%, or casually resampling with a proper weighted method, it isn't too bad, but just resampling this a little bit with Nearest Neighbor, and it's artifact city.

What sensor are we looking at here, and what is the resampling history of the 2.9MP image presented?

With electronic camera displays, it should be possible to have a feature for manual focus that accomplishes this by taking 9 (3x3) small sensor areas and showing them at 100% or 200% on the display. Oh the tragedy, of not having open-source camera functionality.

With some lenses, because of problems like distortions, flat field, lower resolution at corners, improperly aligned mount, etc. the old mirror trick (Zig-Align, Hasselblad's linear mirror) IMHO is still valuable for critical tests and applications.

Good question!

So the photo was taken with the Canon 6D2 + Tamron 35-150 / 2.8 at 50mm f/5.6 1/500 ISO 100. I use PhotoLab 6 for the RAW conversion, using Deep Prime noise filtering at a setting of 10 (out of 100), lens corrections enabled, with no additional sharpening. I then used Topaz Denoise (forget the settings -- varies wildly from photo to photo, but probably something like 15/15 (out of 100) for NR/sharpening. Lastly, I use Arles (doubt anyone here knows of it -- it's old and not supported anymore) to downsample and add the frame (it's at this stage that the EXIF is stripped -- I am not sure if there's a setting that allows me to keep the EXIF intact -- same if I were to use IrfanView, by the way). For the sharpening in the downsampling, I typically use "Hermite" at the lowest setting but sometimes use "Mitchell", either at the lowest setting or one up.

It passed. Further testing, however, is always required. 😉

Then again, maybe it was just wall paper over a sloppy cement job -- I didn't cross the street to verify. 😁

So this is a 33.333% downsample? Looks like a possible 3x3 bin.

Very close -- the fullsize is 5996 pixels wide and the downsampled photo is 2044 pixels wide (not including the frame): 2045 / 5096 = 34.089%. Still might be using a 3x3 bin. Here's a 100% crop:

Interesting the need for sharpening when down-sampling. Bart van der Wolf warns of possible aliasing caused by just down-sampling:

kronometric.org/phot/processing/Down%20sampling%20methods.htm

You never want to sharpen before downsampling. Increases the energy of aliased frequencies. Sharpen afterwards.

I always process photos for the fullsize, so sharpening is inevitably applied before downsampled for web display.

Many of us apply 'capture' sharpening to our fullsize images. And that's ok as long as only true 'capture' sharpening is applied, say by light deconvolution - and appropriate filtering is performed before/during downsizing. I find that this stuff is a little easier to think about in the frequency domain:

www.strollswithmydog.com/downsizing-algorithms-effects-on-resolution/

Jack

Kind of reminds me of brickwork on the exterior or a building.