4

I am working on some pre-processing for lung CT images. I see a nice tutorial in here. Two of them are normalization and zero centering. I wonder what is the difference between these two steps? If one normalizes the image pixel values to range [0, 1], is it have a benefit to also do zero centering? Is there a difference if I do normalization first or zero centering first?

Shayan Shafiq
  • 1,008
  • 4
  • 13
  • 24
Hunar
  • 1,197
  • 2
  • 11
  • 33

1 Answers1

3

what is the difference between these two steps?

In that specific Notebook that you linked, normalization means: shrink a numerical distribution in the [0,1] interval. It is commonly referred to as Min-Max Scaling. Shrinking the distribution in the [0,1] interval moves its mean somewhere between 0 and 1.

Zero-centering instead means: to "shift" the values of the distribution so that its mean is equal to 0.

--

if one normalize the image pixel values to range [0, 1], is it have a benefit to also do zero centering?

Zero centering variables improves ML models performance. Activation functions tend to be very responsive to weights' changes especially around zero. I found some good explanations here.

--

Is there a difference if I do normalization first or zero centering first?

The order matters, since "normalization" would move the mean away from zero.

Leevo
  • 6,445
  • 3
  • 18
  • 52