NSXType-R said:
You can use MATLAB for photos too?

Of course. An image is just three (or four) matrices of numbers.

Basic MATLAB has some image functionality, but they also sell an Image Processing Toolbox with more bells and whistles. These are not really aimed at photographers, and don't have a nice user-friendly interface or workflow.

But, if you have the time, it is great for things you can't do otherwise.

For instance, I wanted to see how much my camera and tripod moved if I took a series of frequent time-lapse photos. So I wrote some MATLAB code to do the following analysis of a bunch of RAW images in a directory.

- Present the first image in a gui and select a small area for analysis (one with strong edges in it).
- Crop every photo to that area.
- Convert photos to greyscale.
- Do a histogram stretch to increase contrast.
- Do a 2D cross-correlation comparing each photo with the first photo.

This showed that the camera moved by one or two pixels when taking 100 shots at 10 second intervals.

You can use MATLAB for photos too?

Up until now, I've just used the default down-sampling in GIMP or whatever image-editing software I am using, assuming it takes care of things properly.

As a D800E pre-owner, I probably need to think more about downsampling. I use MATLAB for my work (not on images), and it has a lot of tools and stuff I can play with to experiment.

With the release of the D800/800E there has been a lot of mention of downsampling to improve image quality. So how do NR enthusiasts do the downsampling? What is best practice?