Add a way to drop rows / columns of un-dividable 2D arrays using 'block_reduce'

[CallShaul]

A suggestion:

Adding a boolean option for reducing output 1D / 2D arrays size, in case the down-sampling integer factor don't match perfectly (block_size).

[rfezzani]

Hi @CallShaul, thank you for this suggestion. Can you illustrate the issue that you are solving with some lines of code?
An example can help us better understand the problem 😉

[CallShaul]

import skimage.measure as skm
import numpy as np

ds_block=(10, 10)
vid = -uint8 video data file- [x, y, time]
img = skm.block_reduce(vid[:, :, k], ds_block, func=np.mean)

As you can see in the image below the lowest row has a lower gray level values since the image wasn't perfectly divided by ds_block.

I've solved the issue slicing the extra rows / columns:
spare_rows = np.mod(vid.shape[0], ds_block[0])
spare_cols = np.mod(vid.shape[1], ds_block[1])
vid = vid[:vid.shape[0] - spare_rows]
vid = vid[:, :vid.shape[1] - spare_cols]

it will be more elegant if the block_reduce function will be able to receive a Boolean argument mentioning whether we want those lines removed or maybe even up-sample by average or something like that to fit a perfect block division.

I hope I was clear enough, I'm available for any further discussion, thanks a lot!

Shaul.