Synthetic aperture radar (SAR), geophysical and seismic imaging, ultrasound imaging, optical coherence tomography (OCT), and many forms of astronomical imaging are examples of coherent imaging systems. In practice, the acquired phase data are noisy due to imprecise knowledge of the motion of the sensor, or due to the properties of the medium. As a consequence, the produced imagery is improperly focused. Existing techniques for correcting the imagery, referred to as autofocus, are ill-equipped to handle the challenges of modern high-resolution systems. The availability of well-focused ultra-high-resolution SAR imagery is crucial for antiterrorism efforts. Improved focusing may also lead to breakthroughs in medical imaging, such as ultrasound imaging of the human breast.
The investigators develop a new generation of autofocus algorithms that will enable improved focusing in a variety of coherent imaging modalities. Primarily, the SAR autofocus problem is considered because of its national importance in all-weather surveillance, and also because the SAR field is highly developed. The research will focus on the following approaches: (i) optimization strategies that optimize image sharpness metrics, (ii) multiresolution vector space methods that exploit the structure of the phase error, and (iii) multichannel blind deconvolution techniques that exploit the redundancy of the blurring operation on each row of the image. These approaches motivate the development of a unified framework, in which all of the available assumptions are used in a robust restoration procedure. The implementation of the methodology in practical SAR systems will be emphasized. Finally, the developed autofocus methodology will be extended and applied to ultrasound, OCT, and passive radar imaging.
