Multiband deblurring for fluid lens cameras.

摘要

Unique image processing challenges are produced by the Fluid Lens Camera System. Over traditional glass lens systems, better miniaturization potential and fixed length lens while zooming are unique abilities offered by the fluid lens. Non-uniform blurring for each color plane of the image is also caused by the fluid in the lens. A sharp green color plane and blurred blue and red color planes are also produced by this fluid. For natural and medical images, the edges in the green and blue color planes are similar. In this work, the sharpness of the blurred color planes is improved by sharing edge information between color planes. Avoiding shading artifacts while improving edge quality is the goal of this work. Several algorithms are discussed: a wavelet-based algorithm, a contourlet-based algorithm, a Support Vector Regression algorithm, and an Adaboost classification algorithm. In each algorithm, the strengths of the previous algorithm are built upon in order to improve results. A major advantage of these methods is that shading and edge information is combined without using a complicated point spread function. While the focus of this dissertation is on using the green color plane to improve the blue color plane, the same algorithms could be applied to the red color plane as well. Infrared imaging, medical image overlaying, satellite mapping, and remote sensing are all multiband system with high edge correlation where this work could be applied.