| 摘要 |
Estimating the spatial frequency response (SFR) of an image acquisition system or device from its output images is accomplished by a image processing system and method thereof even when the original scene content cannot be carefully controlled as in the laboratory method. The image processing system methodology assumes that the scene contains specified reference signals (e.g., step edges), but that the reference signals' location, orientation, contrast, mean gray level, and spatial extent in the corresponding digitized image(s) are unknown. Reference signal responses are searched for and extracted using a paradigm, culminating in a set of candidate responses which survive all the tests by scoring highest on the goodness metrics. These surviving candidate responses are each fit, alone or in groups, to ideal responses that would occur if the given reference signal was passed through likely system models. The SFR of the system which best fits each available candidate response is an estimate of the input imaging device SFR. Each candidate response therefore has three computed quantities associated with it: 1) the SFR corresponding to the system producing the best fit, 2) a goodness-of-fit value from the fitting procedure, and 3) the goodness score achieved by the candidate during the initial identification and pruning procedure. A centroid in this 3D parameter space, weighted heavily toward candidates with high integrated SFR, is used to choose the final SFR system estimate. Step edges may be used as the reference signal, but the system is not limited thereto. |
