Higher image resolution when scanning in v3

Currently scanning in v3 captures images at 0.5 megapixel (832x624) which is every 4th pixel from the camera. Full Res 8 megapxel (3280x2464).

The question is when does increasing the camera resolution stop giving a benefit as we are beyond the optical resolution.

This can be calculated, but there are some caveats. The optical resolution limit is of order the spacing for a 1MP image.

However, in capturing a 0.5MP image, each colour channel is spaces over an equivalent 0.125MP image. Even for a capture of a 2MP image the each channel is a 0.5MP image. To get all channels above the optical limit a full 8MP image where each channel is 2MP is required.

It is also worth considering that collecting from more pixels should increase signal to noise.

Data from a 20x objective looking at a Stomach section

These are not the full images. These are sections of the images zoomed in on fine detail within the image:

Capture resolution and scaling	Image
8MP
8MP downsampled to 2MP
2MP
0.5MP

Conclusion

To me this implies that we do need to capture at 8MP, but we can downsample to 2MP after capture.

Note that the done with default ImageMagick settings. For downsampling an image without transparency that should be Lanczos downsampling.

Extra:

Capture resolution and scaling	Image
8MP downsampled to 1.1MP
8MP downsampled to 0.5MP
8MP downsampled to 0.5MP using Box rather than Lanczos

I would argue that if space is an issue downsampling to ~1MP is possibly fine, but 0.5MP is too low.

The Box resample to 2MP is very fast and performs well. Note that Box has a number of weird antialiasing effects if downsampling by an arbitrary amount. However if binning (reducing an integer number of pixels into 1 pixel) it shouldn't have these effects. So converting from 8MP to 2MP (2x2px becomes 1 px) is OK converting from 8MP to 1.1MP (2.66x2.66px becomes 1px) wouldn't perform well.