CT from Motion: Volumetric Capture of Moving Shapes with X-rays and Videos

Julien Pansiot and Edmond Boyer

BMVC 2017

CT from Motion


In this paper, we consider the capture of dense volumetric X-ray attenuation models of non-rigidly moving samples. Traditional 3D medical imaging apparatus, eg. CT or MRI, do not easily adapt to shapes that deform significantly such as a moving hand. We propose an approach that simultaneously recovers dense volumetric shape and motion information by combining video and X-ray modalities. Multiple colour images are captured to track shape surfaces while a single X-ray device is used to infer inner attenuations. The approach does not assume prior models which makes it versatile and easy to generalise over different shapes. Results on synthetic and real-life data are presented that demonstrate the approach feasibility with a limited number of X-ray views. The resulting dense 4D attenuation data provides unprecedented insights for motion analysis.


CT from Motion
CT from Motion

Bibtex reference

	author     =  {Julien Pansiot and Edmond Boyer},
	title      =  {{CT} from Motion: Volumetric Capture of Moving Shapes with {X}-rays and Videos},
	booktitle  =  {British Machine Vision Conference (BMVC)},
	year       =  2017,
	month      =  Sep,
	address    =  {London},
	eprint     =  "http://julien.pansiot.org/papers/2017_Pansiot_BMVC_CTfM.pdf",
	video      =  "http://julien.pansiot.org/suppl/2017_Pansiot_BMVC_CTfM.mp4"
	url        =  "https://bmvc2017.london/programme-1/"