Hi all,
I am running an fMRI experiment. During the task, I'm showing visual stimuli (black and white) to my participants. To exclude any alternative explanation of the results, I need to make sure the stimuli don't have any low-vision perceptual difference.
So the question is: How can I control brightness and contrast of a set of stimuli? Is there any standard procedure to make sure that every stimulus has the same average brightness and contrast?
Thanks,
Andrea
Andrea,
if your stimuli are relatively simple shapes (letters of the alphabet, geometric shapes like circles, triangles...) control of brightness and contrast is fairly straightforward. Here are the steps:
1) use a photometer to measure the Luminance of 20 or so RGB combinations ranging from say 35, 45 up to 255, in increments of 10. Because your stimuli are BW, your RGB combinations would then be 1) R=G=B=35, 2) R=G=B=45, 3) R=G=B=55, ....23) R=G=B=255
2) statistically model your photometric data with a nonlinear function (exponential functions work well) -- you should get a highly predictive functional fit (Rsq > .9).
3) with this function, you can design the RGB values of your stimuli and stimuli-backgrounds to achieve whatever luminance contrast and brightness (luminance) values you desire.
That's it! -- but here are a few fine points to consider:
a) if your stimuli are complex images with fine detail -- e.g., a BW image of a natural scene or a persons face, then brightness and contrast will vary substantially across the image (depending upon what location within the image the observer is focusing upon at a given moment). So in this case, it doesn't make a lot of sense to claim you have "controlled" contrast and brightness. If on the other hand you are presenting a single BW stimuli, with a single RGB value across its area, and the background of the stimuli is one single RGB value, then to claim you have control over contrast and brightness is more believable
b) regarding the photometric calibration of your display screen (#1 above). you could start at RGB = 0 and go to RGB=255 in increments of 1, but that's a lot of work, and in my experience you don't get a lot more predictive power by doing so. Most displays produce very little light at low RGB levels, and most low-midgrade photometers aren't very accurate or precise measuring low light levels (say less than 10 cd/m2). So starting around 30, 40 or so all the way to 255 (in increments of 10) gives you a good quality function
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Signal Processing
- Image Processing