Hi ! Thanks for the great tool and amazing breath of data from the Allen Brain Observatory.
I was wondering which specific normalization method was used to normalize the contrast in the natural movie clips? And did this normalization happen over the entire clip (ie via combination and normalization of all frames together) or over individual frames?
Thank you!
Dahlia
Hi Dahlia,
I’m glad you are finding it useful!
Contrast normalization was computed via the root-mean-square method: sqrt( mean( x - x_mean )**2 )/ x_mean , where x is a single frame of the image, x_mean is the mean luminance over the frame, and the mean is taken over the pixels in one frame. Each frame is normalized to 60% contrast via this method. There were some error we found after the fact due to cast values to integers, so it isn’t a perfect 60%, but this was a fairly minor difference.
Saskia
Hi Saskia,
Great, thank you for the helpful information!
Just a clarification question – x is the value of a single pixel or of the frame? Also, the movie clip stores in the allensdk is before this normalization, correct?
Thank you!
Dahlia
Hi Dahlia,
x is the array of values for all the pixels in the frame.
The movies arrays available via the SDK are after the normalization.
Saskia