For this particular sample, though, there was a lot of background noise with low pixel values. If I move the slider far enough to erase those, I also start erasing important details in the text.
So we have to go at the noise a different way. One way to do this is using a "high-pass filter." A high-pass filter processes the image, mapping out where pixel-to-pixel variation is high across a certain distance. You tell it a pixel scale to work with. Here I'm trying 3 pixels.
I should have mentioned, copy and paste the original image (not the levels adjusted one) into a new layer before doing the high-pass filter, because the output will be a whole layer that has been filtered this way. That's basically a noise image; we want to subtract it from the original.
So now I have the original image in one layer and the high-pass filtered image in the layer on top of it. Then I set the blending mode of the high-pass filtered image to "divide." This is a lot better, though still not perfect.
