Read Me
This is the python interface for the MIA library. It uses numpy to
represent images, and currently, filters and image registration is
supported.
Compiling the interface:
The build system utilizes python distutils. To compile the interface
the compiler must support the -std=c++11 compiler flag.
Building the interface then build down to running
python setup.py build
Note, that you wil need mia >= 2.0.10 to be installed.
Using the python interface:
Given that MyImage is a 2D or 3D numpy array of scalar values, one may
run a filter chain like
import mia
MyFilteredImage = mia.filter(MyImage, ["filter1:param1=a",
"filter2:param1=b,param2=c", ...])
For the availabe filters see the user reference of MIA.
Similarly, the image registration can be run with images Moving and Reference,
using a spline based transformation with a coefficient rate of 5 pixels, 2
multi-resolution levels and an nlopt based optimizer:
import mia
Registered = mia.register_images(src=Moving, ref=Reference,
transform="spline:rate=5", cost=["image:cost=ssd", "divcurl:weight=10.0"],
mglevels=2, optimizer="nlopt:opt=ld-var1,xtola=0.001,ftolr=0.001,maxiter=300")
The result image is stored in Registered. Currently, the transformation is lost.
A simple segmentation program for the segmentation of myocardial perfusion images
is available as mia-myocard-segtool. It requires Python-WxWidgets to be installed.
The image series is expected to be given as a XML file that can be created by using
mia-2dseries2sets.