Motion Compensated Image Reconstruction
Prof. Jeff Fessler, EECS
Image reconstruction of moving objects (such as breathing patients) is challenging due to inconsistencies between measurements acquired at different phases of the motion. Compensating for motion during image reconstruction requires tools similar to those used in nonrigid image registration. In the first part of this talk I will discuss an approach for nonrigid image registration based on B-spline deformation models. The key feature of this approach is that it provides a simple way to ensure that the estimated deformation is invertible (diffeomorphic), using sufficient conditions based on the inverse function theorem. This constraint is important for the registration to be physically plausible.
In the second part of the talk I will describe a few approaches for using this type of image registration tool in the context of image reconstruction of moving objects. Although the motivation of this work is medical imaging, the ideas are also relevant to the problem of recovering a high-resolution image from a low-resolution video sequence of moving objects ("super resolution"). (This is work done mostly by EE:Systems graduate student Se Young Chun and by EE:Systems graduate Matt Jacobson.)
For Prof. Fessler's biography, click here .
Thursday, October 2