Having dealt with and printed a lot of very large images, e.g., 60k x 60k pixels, I have been on the lookout for image processing software that never decompresses the entire image into ram, but instead works on blocks or scan lines or blocks of scan lines, but stays in constant memory and streams to and from disk. For example, the ImageMagick fork GraphicsMagick does a much better job of this than ImageMagick. What other software is out there that can handle these kinds of images?
The key is not to store it in raster form in RAM. Either tiles (like GIMP) or I prefer Z-ordering. Then a user can zoom in and pan around easily - you let the system swap and it won't be bad at all. If they zoom out though, you probably want to store MIP maps of it.
Swap works well for this as long as your data has good locality. huge raster images don't.
But no, I'm not aware of any software that handles stuff like that well - except the GIMPs tiling, but that's not going to help when zoomed out.
I definitely want to avoid swap at all costs and find things that are designed to tile & stream instead. The difference between GraphicsMagick resizing an image by streaming and ImageMagick resizing an image that hits swap is orders of magnitude - seconds versus hours.
You divide the image into quarters and store each quarter as a continuous block of memory. Do this recursively.
Normally we'd index into the pixel data using pixel[x,y]. You can get Z-ordering by using pixel[interleave(x,y)] where the function interleave(x,y) interleaves the bits of the two parameters.
This works fantastically well when the image is a square power of two, and gets terrible when it's one pixel high and really wide. I think a combination of using square tiles where each one is Z-ordered is probably a useful combination.
For my ray tracer I use a single counter to scan all the pixels in an image. I feed the counter into a "deinterleave" function to split it into an X and Y coordinate before shooting a ray. That way the image is rendered in Z-order. That means better cache behavior from ray to ray and resulted in a 7-9 percent speedup from just this one thing.
Once you have data coherence, swapping is not a big deal either in applications where you're zoomed in.
> You can get Z-ordering by using pixel[interleave(x,y)] where the function interleave(x,y) interleaves the bits of the two parameters.
That is pretty cool (also the raytracing application), but most surprising to me, is I remember the interleave operator from reading the INTERCAL[0] docs ... I never even considered the function could actually be applied to something useful :-)
[0] one of the early esoteric programming languages, designed to be "most unlike any other language". It also features a "COME FROM" statement (because GOTO is considered harmful), which iirc also actually has a parallel in some modern programming paradigm.
Nuke works in scanlines like this, and can process a whole tree of operations only loading the input lines necessary for the current output row. The SDK docs explain the architecture somewhat: https://www.thefoundry.co.uk/products/nuke/developers/90/ndk...
