Mark Rivers

>Since it does seem from your log output that the entropy value is decreasing, I suggest first trying a smaller tol value. This should make SciPy search longer. If that...

The paper you cite is what I used in my IDL code. The paper explains how to compute the entropy. It is the sum of H*log(H) where H is the...

I tested my idea that increasing the number of histogram bins from 64 to 10000 would fix the problem. I was wrong, it did not help. I then changed my...

I modified the program in the previous comment as follows, adding the call to tomopy.normalize_bg. ```python logging.info('Normalizing ...') norm = tomopy.normalize(proj, flat, dark) logging.info('Secondary normalize to air ...') norm =...

I tried changing the padding to constant mode with a value of 1 as follows: ``` norm = tomopy.misc.morph.pad(norm, axis=2, npad=npad, mode='constant', constant_values=1) ``` These are the results: Method |...

Here is a comparison of 0-180, entropy, and Vo both with and without padding. Here I also list the execution time. The times listed are the times to center both...

> A quick feedback after I read some of the comments. There are few differences from the IDL entropy based centering code that I remember. First, the ring correction is...

I am trying to track down why tomopy and IDL produce such different results for centering based on entropy. I first wrote a little program that reads in the same...

I realized that in the above test I was setting mask=False in tomopy.find_center(). That is not the default, and not what I was using in the tests above yesterday. I...

Yesterday I showed that tomopy.find_center() failed badly if air normalization was applied. I just changed the test program above to add the air normalization as follows: ``` logging.info('Reading data ...')...