python - Fastest way to convert ubyte [0, 255] array to float array [-0.5, +0.5] with NumPy -

The question is in the title and it is very straightforward.

I have a file from F I am reading the ubyte array:

  arr = numpy.fromfile (F, '& gt; u1', size * rows * cols Currently, I am rearranging the data in 3 passes, as follows:      <     
  arr = images.astype (float) arr - = max_value / 2.0 arr / = max_value  

Since The array is somewhat larger, it takes an important part of each other.
It would be nice if I could do this through data 1 or 2 near Because I think it will be faster.

Do I have to reduce the number of passes for a "composite" vector?
or, do I have to speed up this speed Another way?

itemprop = "text">

I did:

  ar = ar - 255/2 r * * = 1./255  

looks fast :)

No, I have finished it, this is my system It is almost twice as fast as it seems that ar = ar - 255/2 In addition to decreasing and type conversion on the fly, it does not optimize the segmentation with a scalar: it is faster to split a bunch of times many times on the array and then on the array. Although additional floating-point operation round-of-error may increase.

As stated in the observations, it is actually a fast yet easy way, to achieve it. This is another factor on my system, which is faster than two, but mostly due to the use of numexpr multiple cores and not so much that it only passes a single pass on the array code: < / P>

  import numexpr ar = numexpr.evaluate ('(- 255.0 / 2.0) / 255.0')