Affine Transformation Pixel Intensity

Hello Fantastic Folks,

I have performed an affine transformation to register two images and am happy with the how the transformation performed. These two images were taken at two different angles and I would like to measure the mean pixel intensity of an ROI in each image to make a statement about how intensity changes at different angles of light. I wonder if anyone forsees an issue with using an affine transformation followed by pixel intensity measures. The method seems like a better option than warping transformations and a rigid transformation does not allow for accurate registration. Any input would be greatly appreciated.

Thank you!


Hi @shnmchlgrffn,

I like that you’re thinking about this and asking this question :slight_smile:

If you want to be extra safe, the alternative would be to transform the ROI instead of the image
and compute your stats in the original image space - a fine choice if you ask me.

However, I expect the mean intensity over a big enough ROI shouldn’t have any systematic issues.
How are you interpolating the image? Most interpolation methods (like bi-/tri-linear) end up effectively low-pass filtering the image, which could cause problems if you care about min/max intensities.


1 Like

I am using bilinear interpolation but the scaling is minimal. The majority of the transformation consists of shearing and translation. My ROIs are 5x10 pixels.