Mechanical image acquisition systems (such as medical C-arms) frequently accumulate geometrical errors which must be calibrated out using a calibration phantom. A more frequent regime of system calibration implies a less frequent use of the C-arm for clinical applications.The present application proposes to identify common biases between the acquired projection frame sequences from the same mechanical image acquisition system in first and second acquisitions, and to compare this to expected calibration data of the mechanical image acquisition system to generate frame deviation measures. If a resemblance between the first and second sequences of frame deviation measures is obtained, one or more calibration actions are performed (such as alerting the user that calibration should be provided, and/or automatically correcting for the geometry deviation).
