A raw sensor is clearly not easy to just hook up to a computer.
For example look at the top end Raspberrypi sensor. It's a pathetic 12MP. That's like a ten year old phone or so?
I think the processing is also not to be entirely dismissed. There is frame stacking that extends the dynamic range and there is compression and other complex DSP going on that is necessary (b/c 50MP of raw pixel data is a ton of raw data to pull off the sensor). Realistically you probably can only do some of that in software
You can do all of these things in software, and it is done. It’s important thing to have control over the process so you can get quantitative data at the other end, and not just a pretty picture. Also, noise should not be discounted as a very good reason to use lower megapixel sensors. If you want a pretty picture, by all means use a cellphone, but you can’t reality use or trust the result for many scientific purposes.
I think this is just not realistic. "Pretty pictures" are actually more important even in science. The vast majority of the time you're not using pixel values or exact color characteristics in a scientific sense. You just want a clear high res image of what you're looking at so that you can ID the pollen, plankton or whatever it is. The algorithms in phone cameras are some of the most advances available. Sure you could in theory reproduce it in software.. but realistically there are no opensource code bases that can recreate the same level of dynamic range that a high end phone company software stack does
I take a pic with my iPhone and I can ID the things on my slide much better than with the color accurate high end Olympus scientific sensor. And in the end that's what matters most
High pixel count cameras aren't that useful in microscopy. Big low noise pixels are better. For every sensor there is already a breakout board with USB. Mostly intended for machine vision.
For example look at the top end Raspberrypi sensor. It's a pathetic 12MP. That's like a ten year old phone or so?
I think the processing is also not to be entirely dismissed. There is frame stacking that extends the dynamic range and there is compression and other complex DSP going on that is necessary (b/c 50MP of raw pixel data is a ton of raw data to pull off the sensor). Realistically you probably can only do some of that in software