Big vs small pixels with the ISO 12233 target

Yesterday, I posted images of Bruce Lindbloom’s imaginary desk with simulated cameras of 10, 5, 2.5, and 1.25 micron pixel pitch, at ISOs of 100,800, and 3200, with and without a 4-way beam-splitting antialiasing filter with a null at 1.33 times the Nyquist frequency.

Today I’ll post a similar 24-image suite with a lowered-contrast version of the ISO 12233 chart as the target image. All are resampled to the resolution of the finest-pitch sensor in sRGB using Lanczos-3.

Here are TIFF files:

http://www.kasson.com/ll/ISO12233TIFF.zip

ANd here are the JPEG files:

http://www.kasson.com/ll/ISO12233JPEG.zip

My comments follow:

Without the AA filter, there is an inconsequential difference between the resolution of the 1.25 micron and 2.5 micron images.

1.25 um, ISO 100:

2.5 um, ISO 100:

With the AA filter, that small difference becomes even smaller.

1.25 um, ISO 100:

2.5 um, ISO 100:

The ISO 3200 images at 1.25 and 2.5 micron pitches are unacceptably noisy. This low contrast target with lots of light gray space is a tough test for noise.

1.25 um, ISO 3200:

2.5 um, ISO 3200:

There is no false color at 1.25 um, and virtually none at 2.5 um. There is some at 5 um, and a lot at 10 um. This target is quite good at generating false color.

5 um, no AA filter, ISO 100:

10 um, no AA filter, ISO 100:

There is a substantial improvement in resolution going from 10 to 5 um pitch. There is a smaller, but significant improvement in resolution going from 5 to 2.5 um pitch. There is almost no improvement going from 2.5 to 1.25 um pitch.

The AA filter does nothing good and not much bad at 1.25 um. The AA filter adversely affects resolution a tiny bit at 2.5 um, and removes the tiny bit of false color that exists in the no-AA-filter 2.5 um image.

The AA filter removes most, but not all of the false color at 5 um, but also adversely affects resolution. My take is that whether you’d want it or not is dependent on the subject.

5 um, AA filter, ISO 100:

The AA filter removes most, but not all of the false color at 10 um, but also adversely affects resolution. My take is that you’d want it for most subjects.

10 um, AA filter, ISO 100:

2.5 micron pixels are looking pretty good.