| 摘要 |
The invention relates to methods for determining the transfer function of a signal-processing system, which do not presuppose a known input signal. The methods are based on two representations I1(x) and I2(x) of an object which have been produced by the system from differently scaled input signals originating from the object, or from a representation I1(x) of a first object and from a representation I2(x) of an object which is geometrically similar thereto but has been scaled differently. The representations are either given or are produced at the start of the method. According to the invention, the representations are transformed into a working space, and sections which relate to the same region of the object are selected in each case. The quotient of the functions corresponding to these two sections in the working space from which the unknown input signal comes makes it possible to clearly determine the transfer function sought. Different methods are stated for this determination. The method can be used, in particular, to improve the images from electron microscopes for which there are no suitable test structures for determining the transfer function. |
