Home

Welcome to the isetcalibrate wiki!

Notes from David Brainard that will get me started.

i) The PLoSOne paper (Tkacik_etal_11.pdf) gives a pretty complete picture of how we calibrated a camera for acquisition of a natural image database. The code that does the calibration described in this paper is on github, and there is a README file at the repository that gives some information. I dusted this off last summer for an undergraduate who wanted to work with some additional images that aren't in the public database. Much of it is probably idiosyncratic to the particular set of measurements we had, but some may be of use. But the code might be useful for teaching/training if someone wanted to follow along with the description in the paper. I'm happy to provide all the raw data. There is a ToolboxToolbox configuration file that sets up preferences to point at where the data live, so I think it would be pretty easy to get this to run.

The code above interoperates with some routines in the BrainardLabToolbox/D70Toolbox.

There is a funny story about the title of this paper, which I'll try to remember to tell you when you're here.

ii) The second paper (Bunya_etal_13) uses an image of a Macbeth color checker to estimate color correction for a camera, so that we could standardize the clinical images across sites where the illumination varied site to site, and estimate colorimetric properties of the images. There was also image to image correction based on an index card that subjects held up next to their eyes. The methods here were constrained in interesting ways be requirements of a clinical trial. I have the code for this, which I'd be happy to share. At one point I packaged it up for someone at Duke to use, so I think I could dig up a version that was moderately well documented.

iii) Remember dcraw (https://www.cybercom.net/~dcoffin/dcraw/) to extract raw images from the cameras in both of these projects. One very important caveat is that from version to version, it changes its output conventions in surprising and undocumented ways. So you need to pick a version and lock it down between calibration and when you use the calibration information on acquired images. - Suggests dockerizing!!! -- DHB has c source code for a number of versions of this program.