Method for processing images using object signals to estimate transfer functions of optical fibers
A method for processing images acquired by image detectors with non-uniform transfer functions and irregular spatial locations includes the steps of accumulating data from multiple images, defining an affinity graph which edges define pairs of detectors that measure related signal, performing statistical analysis on the accumulated data with respect to each pair of detectors, and solving a system of equations constructed from the results of the statistical analysis to estimate each detector transfer function, a set of solutions to the system of equations comprising a calibration of an imaging system.
1. A method for processing images acquired by an imaging system comprising image detectors with non-uniform transfer functions and irregular spatial locations, the method comprising:
accumulating data from multiple images;
defining an affinity graph which edges define pairs of detectors that measure related signal;
performing statistical analysis on the accumulated data with respect to each pair of detectors; and
solving a system of equations constructed from the results of the statistical analysis to estimate each detector transfer function, a set of solutions to the system of equations comprising a calibration of the imaging system.
2. The method according to claim 1 , wherein, the statistical analysis comprises regressions.
3. The method according to claim 2 , wherein calibration is performed as soon as a quantile of a correlation coefficient associated to regression on the accumulated data reaches a threshold.
4. The method according to claim 1 , wherein a time-varying adaptive calibration is performed by changing a temporal window.
5. The method according to claim 2 , wherein calibration is performed continuously as soon as a quantile of a correlation coefficient associated to regression on the accumulated data reaches a threshold.
6. The method according to claim 1 , wherein the accumulated data is composed from successive images for which the correlation coefficient computed between said successive images reaches a threshold.
7. The method according to claim 5 , wherein different parameters of the detector transfer functions are updated at different times depending on the content of the accumulated data.
8. The method according to claim 1 , further comprising acquiring a background image during initialization of the calibration of the imaging system.
9. The method according to claim 1 , wherein each calibration is saved and restored for ulterior initializations.
10. The method according to claim 1 , wherein the affinity graph is constructed from a neighborhood information, in particular, based on the Delaunay triangulation of the spatial position of the detectors.
11. The method according to claim 1 , further comprising using the calibration for image reconstruction.
12. The method according to claim 1 , wherein the data is pre-processed or filtered prior to performing the statistical analysis on the linked detectors.
13. The method according to claim 1 , wherein computations associated with solving the system of equations are stabilized by using prior information, goodness of fit measures from regressions, or robust computation methods.
14. The method according to claim 4 , wherein computations associated with solving the system of equations are stabilized by using prior information on temporal evolution or temporal smoothness constraints.
15. The method according to claim 8 , wherein a model of a temporal background evolution is used to update an initial estimation of the background image.
16. The method according to claim 1 , wherein the calibration is computed off-line or in lagged time.
17. The method according to claim 16 , wherein, several different sequences of images are combined to compute the calibration.