# Proper way to flip using AffineTransform with SimpleITK
**URL:** https://discourse.itk.org/t/proper-way-to-flip-using-affinetransform-with-simpleitk/3898
**Category:** Beginner Questions
**Tags:** python, simpleitk
**Created:** [February 24, 2021, 12:51pm UTC](https://discourse.itk.org/t/proper-way-to-flip-using-affinetransform-with-simpleitk/3898 "2021-02-24T12:51:06Z")
**Posts on this page:** 3
**Page:** 1
### Author:  [@bobdark](https://discourse.itk.org/u/bobdark)
#### Post date: [February 24, 2021, 12:51pm UTC](https://discourse.itk.org/t/proper-way-to-flip-using-affinetransform-with-simpleitk/3898/1 "2021-02-24T12:51:06Z")
I’m trying to flip a 3d image using AffineTransform (I have to support multiple possible modifications to a 3d image, like rotation and flipping, so I’m trying to represent them as matrices) and it seems that when I apply a flipping affine transform [1,0,0,0,1,0,0,0,-1] - to flip around Z axis, the image direction is not modified. If I use sitk.Flip([False, False, True]) instead, the resulting image direction is correct (z axis is -1). Do I need to update the image direction manually after applying affine transform?
```
import SimpleITK as sitk
image = sitk.Image(3, 3, 3, sitk.sitkInt16)
image.SetDirection([1.0,0.0,0.0,0.0,1.0,0.0,0.0,0.0,1.0])
image.SetPixel(0,0,0, 5)
image.SetPixel(0,1,0, 5)
image.SetPixel(0,2,0, 4)
image.SetPixel(1,0,0, 5)
image.SetPixel(1,1,0, 5)
image.SetPixel(1,2,0, 4)
image.SetPixel(2,0,0, 5)
image.SetPixel(2,1,0, 5)
image.SetPixel(2,2,0, 4)
image.SetPixel(0,0,1, 6)
image.SetPixel(0,1,1, 6)
image.SetPixel(0,2,1, 6)
image.SetPixel(1,0,1, 6)
image.SetPixel(1,1,1, 6)
image.SetPixel(1,2,1, 6)
image.SetPixel(2,0,1, 6)
image.SetPixel(2,1,1, 6)
image.SetPixel(2,2,1, 6)
image.SetPixel(0,0,2, 10)
image.SetPixel(0,1,2, 10)
image.SetPixel(0,2,2, 10)
image.SetPixel(1,0,2, 10)
image.SetPixel(1,1,2, 10)
image.SetPixel(1,2,2, 10)
image.SetPixel(2,0,2, 10)
image.SetPixel(2,1,2, 10)
image.SetPixel(2,2,2, 10)
image.SetOrigin([0,0,0])
print("--------- Original -------------")
print("Direction: {}".format(image.GetDirection()))
print(sitk.GetArrayFromImage(image))
print("--------- Flip via AffineTransform -------------")
affineTrans = sitk.AffineTransform(3)
affineTrans.SetMatrix([1,0,0,0,1,0,0,0,-1])
affineTrans.SetCenter([1,1,1])
flipped = sitk.Resample(image, affineTrans)
print("Direction: {}".format(flipped.GetDirection()))
print(sitk.GetArrayFromImage(flipped))
print("--------- Flip via sitk.Flip ---------------")
flipped = sitk.Flip(image, [False,False,True])
print("Direction: {}".format(flipped.GetDirection()))
print(sitk.GetArrayFromImage(flipped))
```
results:
--------- Original -------------
Direction: (1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0)
[[[5 5 5]
[5 5 5]
[4 4 4]]
```
[[6 6 6]
[6 6 6]
[6 6 6]]
[[10 10 10]
[10 10 10]
[10 10 10]]]
--------- Flip via AffineTransform -------------
Direction: (1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0)
[[[10 10 10]
[10 10 10]
[10 10 10]]
[[6 6 6]
[6 6 6]
[6 6 6]]
[[5 5 5]
[5 5 5]
[4 4 4]]]
--------- Flip via sitk.Flip ---------------
Direction: (1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, -1.0)
[[[10 10 10]
[10 10 10]
[10 10 10]]
[[6 6 6]
[6 6 6]
[6 6 6]]
[[5 5 5]
[5 5 5]
[4 4 4]]]
```
---
### Author:  [@blowekamp](https://discourse.itk.org/u/blowekamp)
#### Post date: [February 24, 2021, 2:32pm UTC](https://discourse.itk.org/t/proper-way-to-flip-using-affinetransform-with-simpleitk/3898/2 "2021-02-24T14:32:50Z")
You may need to review the SimpleITK’s [Fundamental Concepts on Images](https://simpleitk.readthedocs.io/en/master/fundamentalConcepts.html) to get a clear understanding that the ITK Image’s pixels are entities in physical space not just an indexed array.
The `Flip` filter changes the order of the pixels in the array buffer and the meta-data ( origin and orientation matrix ) such that the physical location of each pixel remains exactly the same. If you use a physical space aware medical image viewer like 3D Slicer to view the “flipped” image before and after it would look exactly the same. However, if you problem the two displayed images you would see different (i,j,k) indexes.
Reimplement the Flip filter with the Resample filter the transformation would be the identity, but the meta-data for the output image would be modifed.
Now if that is understood, the determination for your implementation can be made if you want to change the physical location of your pixels or preserve them.
---
### Author:  [@dzenanz](https://discourse.itk.org/u/dzenanz)
#### Post date: [February 24, 2021, 5:34pm UTC](https://discourse.itk.org/t/proper-way-to-flip-using-affinetransform-with-simpleitk/3898/3 "2021-02-24T17:34:35Z")
This question was also cross-posted to [stack overflow](https://stackoverflow.com/questions/66317658/flip-image-correctly-with-sitk-affine-transform).