Friday, March 5, 2010
The resolution of an imaging system is limited by the wavelength and numerical aperture of the system. Structured illumination imaging has been applied in microscopy to resolve spatial frequencies that conventionally lie outside the passband of the imaging system. The object is imaged with a sinusoidally patterned illumination, rather than the customary uniform illumination. This produces aliased moire patterns which carry high frequency information in the image. These aliasing effects shift the traditionally inaccessible portions of the object's spatial frequencies into the passband of the system making superresolution imaging possible. Structured illumination has also been used to obtain axially sectioned images on similar lines as the superresolution imaging with somewhat modified image processing.
The applications of interest in this thesis are ophthalmoscopy and any moving microscopy application. The human retina is made up of millions of cells. Imaging the human retina in vivo is necessary for the study of structure and function of retinal physiology, and the detection, diagnosis and study of retinal diseases. In vivo retinal imaging with adaptive optics has shown the potential to resolve individual cells non-invasively. Cells such as cones, blood vessels and retinal pigment epithelial cells can now be routinely imaged in the human retina in vivo. The resolution of adaptive optics retinal imaging systems is fundamentally limited by the size of the pupil of the eye. The pupil aperture cannot increase beyond a certain point even with dilating drops. This restricts retinal imaging to structures larger than about 2 microns. In order to image finer features such as rods, foveal cones, ganglion cell axons and dendrites, it is necessary to use superresolution imaging. Similar incentives and restrictions for imaging moving specimens makes superresolution attractive to other forms of microscopy as well.
However, prior work in the area of structured illumination imaging has been restricted to imaging stationary objects with fixed, known phase shifts in the sinusoidally patterned illumination. In this thesis we apply this technique to moving objects. We have modified existing theory and algorithms for structured illumination imaging to accommodate a randomly moving rigid, translating object, such as the human retina in vivo. We show results for phase shift estimation and superresolution in simulation as well as with experimental implementation on a microscope. Further this thesis also explores the implementation of this technique on a widefield, flood illuminated, adaptive optics retinal imaging system. It also addresses some aspects of speckle reduction and image registration for the coherent laser illumination of a moving in vivo retina.
Since many structures in the retina and in microscopy are non-fluorescent, this thesis also explores the coherent case for structured illumination imaging using a laser illumination for non-fluorescent objects. We expand the scope of the theory and algorithms to account for coherence and associated speckle and other artifacts. We show simulation results for this aspect of the project.
We also investigate the application of axial sectioning using structured illumination for moving fluorescent objects. We prove the results for this portion of the thesis using fluorescence images on a microscope.