There is a bug in the way exposure.histogram handles integer arrays. The problem is that the values of the array are shifted because np.bincount can only handle non-negative values. There is however no protection from integer overflow, as in the following example:
>>> from skimage import exposure
>>> exposure.histogram(array([-1,127],dtype=int8))
Traceback (most recent call last):
File "<ipython-input-93-11812a617109>", line 1, in <module>
exposure.histogram(array([-1,127],dtype=int8))
File "C:\Anaconda\lib\site-packages\skimage\exposure\exposure.py", line 60, in histogram
hist = np.bincount(image.ravel() - offset)
ValueError: The first argument of bincount must be non-negativeMy solution would be to view the array as the equivalent unsigned type in the call to bincount.
There is a bug in the way exposure.histogram handles integer arrays. The problem is that the values of the array are shifted because np.bincount can only handle non-negative values. There is however no protection from integer overflow, as in the following example:
My solution would be to view the array as the equivalent unsigned type in the call to bincount.