DIGITAL IMAGE PROCESSING TECHNIQUES FOR PHASE CONTRAST MICROSCOPY (FOURIER OPTICS, OPTICAL SYSTEMS).

摘要

Phase objects are transparent objects which differ from their surround in optical path length, producing images which are spatially modulated in phase, and hence are invisible. Through the use of optical filtering, the phase contrast microscope transforms phase modulation into amplitude modulation, rendering the phase object visible. The transformation is imperfect due to a band-pass effect of the microscope and detector nonlinearities.; Phase contrast image formation can be modeled as a two-dimensional linear system. The phase plate transmittance values (the pupil function) specify the optical transfer function which performs the phase to amplitude modulation. Under ordinary circumstances, the pupil function directly specifies the transfer function. With oblique illumination and an annular phase plate configuration, however, the transfer function has to be numerically computed. An algorithm for this derivation was developed.; Under constraints of small object size and optical path length the phase to amplitude transformation is nearly exact, with little residual phase error in the image. The output from a nonlinear intensity detector can therefore be transformed to amplitude and inverse filtered to remove the band-pass effects of the phase contrast microscope. When these constraints cannot be met, such in blood cell imaging, an inverse filter can be developed to optimize performance for a particular object class. This method of inverse filter derivation is given. Application of the filter as a pre-processing transformation was demonstrated.; The behavior of phase contrast images in response to several object-related variables was studied. From this analysis, a number of image measures unique to phase contrast images were suggested as features for pattern classification. A phase contrast-specific edge detector was also developed and used in feature extraction. Based on a set of phase contrast-specific features the feasibility of phase contrast image classification was demonstrated using a blood cell differentiation problem.; As a result of these investigations, pattern recognition of phase contrast images was found to be feasible. The behavior of phase contrast image formation was studied. Forward and inverse filters were developed to aid in the study of image formation and in the application of feature extraction techniques.