Affine transform xy

[akuchibotla]

In efforts to address #1779

Here's a demo:

import numpy as np
import matplotlib.pyplot as plt
import matplotlib.cm as cm

from skimage import transform

# Generating a square
im = np.zeros((128, 128))
im[32:-32, 32:-32] = 1

# Creating Affine Transform instances with different shearing factors
t1 = transform.AffineTransform(shear=(0, 0.25))
t2 = transform.AffineTransform(shear=(0.75, 0))
t3 = transform.AffineTransform(shear=(0.5, 0.25))

# Transform image with transform objects above
im1 = transform.warp(im, t1)
im2 = transform.warp(im, t2)
im3 = transform.warp(im, t3)

# Display results
fig, (ax1, ax2, ax3, ax4) = plt.subplots(nrows=1, ncols=4, figsize=(12, 3), sharex=True, sharey=True)

ax1.imshow(im, cmap=plt.cm.jet)
ax1.axis('off')
ax1.set_title('Original Image', fontsize=20)

ax2.imshow(im1, cmap=plt.cm.gray)
ax2.axis('off')
ax2.set_title('Shear $y=0.25$', fontsize=20)

ax3.imshow(im2, cmap=plt.cm.gray)
ax3.axis('off')
ax3.set_title('Shear $x=0.75$', fontsize=20)

ax4.imshow(im3, cmap=plt.cm.gray)
ax4.axis('off')
ax4.set_title('Shear $x=0.5 y=0.25$', fontsize=20)

fig.subplots_adjust(wspace=0.02, hspace=0.02, top=0.9,
                    bottom=0.02, left=0.02, right=0.98)

plt.show()

There seems to be some unwanted extra translation occurring though. Any ideas how to tackle this?

[akuchibotla]

Just talked to @stefanv and the translation is expected because of the origin, so looks like it's a non-issue. I'll write up some tests for this.

[stefanv]

The reason the shear goes in a slightly unintuitive direction is because the values are fractions -- if you set it to > 1, then I guess it will look as expected.

[stefanv]

Test both versions of shear, and make sure the pixels end up where they are supposed to. This can be done by, e.g., having a white pixel in an image, warping it, and making sure that white pixel ends up in the right place (use interpolation oder = 0)

[sciunto]

@akuchibotla Did you have a chance to look at the tests?

The PR needs a rebase.

[Ethiral]

Can you tell me the matrix being used in case of only shear?The code shows shear being used only in Y axis

[lagru]

Hey, I think this was addressed in the recently merged PR #6752. I am not sure why this stalled. Nevertheless thank you @akuchibotla for your contribution! Even though it wasn't merged in the end it is appreciated and might have informed later work. :)