Pixelized Sources#
PyAutoLens can reconstruct the light of a strongly lensed source-galaxy using a pixel-grid, using a process
called an Inversion
.
Lets use a Pixelization
to reconstruct the source-galaxy of the image below, noting how complex the lensed source
appears, with multiple rings and clumps of light:

Rectangular Example#
To fit this image with an Inversion
, we first mask the Imaging
object:
mask = al.Mask2D.circular(
shape_native=imaging.shape_native, pixel_scales=imaging.pixel_scales, radius=3.6
)
imaging = imaging.apply_mask(mask=mask_2d)
To reconstruct the source on a pixel-grid, called a mesh, we simply pass it the Mesh
class we want to reconstruct its
light on.
We also pass a Regularization
scheme which applies a smoothness prior on the source reconstruction.
Below, we use a Rectangular
mesh with resolution 40 x 40 and a Constant
regularization scheme:
pixelization = al.Pixelization(
mesh=al.mesh.Rectangular(shape=(40, 40)),
regularization=al.reg.Constant(coefficient=1.0),
)
source_galaxy = al.Galaxy(
redshift=1.0,
pixelization=pixelization
)
To fit the data, we simply pass this source-galaxy into a Tracer
(complete with lens galaxy mass model). The
FitImaging
object will automatically use the source galaxy’s Pixelization
and Regularization
to reconstruct
the lensed source’s light using the Inversion
:
tracer = al.Tracer.from_galaxies(galaxies=[lens_galaxy, source_galaxy])
fit = al.FitImaging(dataset=imaging, tracer=tracer)
Here is what our reconstructed source galaxy looks like:

Note how the source reconstruction is irregular and has multiple clumps of light, these features would be difficult to represent using analytic light profiles!
The source reconstruction can be mapped back to the image-plane, to produce a reconstructed image:

Voronoi Example#
PyAutoLens supports many different pixel-grids. Below, we use a VoronoiMagnification
pixelization, which
defines the source-pixel centres in the image-plane and ray traces them to the source-plane.
The source pixel-grid is therefore adapted to the mass-model magnification pattern, placing more source-pixel in the highly magnified regions of the source-plane.

By inspecting the residual-map, normalized residual-map and chi-squared-map of the FitImaging
object, we can see
how the source reconstruction accurately fits the image of the strong lens:

Wrap-Up#
This was a brief overview of Inverion
’s with PyAutoLens.
There is a lot more to using Inverion
’s then presented here, which is covered in chapters 4 and 5 of
the HowToLens, specifically:
How the source reconstruction calculates the flux-values of the source pixels when it performs the reconsturction.
What exactly regularization is and why it is necessary.
The Bayesian framework employed to choose an appropriate level of smoothing and avoid overfitting noise.
How to perform lens modeling with inversions.
Advanced
Pixelization
andRegularization
schemes that adapt to the source galaxy being reconstructed.