Optical and digital information processing

点击翻译按钮获取中文摘要

Much of our research over recent years has been oriented towards pattern recognition invariant to changes of orientation, scale, 3-D rotations, illumination and other distortions. Here we propose to extend this research and some new approaches. A key concept in many parts of this program is a vector space approach to image representation and processing. The nonlinear locally adaptive correlation filter methods developed to obtain correlation peaks invariant to illumination are being extended to face recognition, and we are extending the vector space approach to change detection in images invariant under local changes of illumination. We are also working in collaboration with industry towards practical applications of some of our previous research using circular harmonic filters for security applications. In collaboration with the defense Department, we will develop the use of correlation techniques along with morphological processing for data mining purposes. Combination of correlation techniques with morphological processing should lead to some more powerful image processing techniques.