Estimation of the head sensitivity function in scanning magnetoresistance microscopy

We apply Bayesian image analysis techniques to a problem in a newly developed scanned probe technology that uses commercial magnetoresistive (MR) record and playback heads as probes to sense magnetic fields. This technology can be used for magnetic imaging and for evaluating playback and record processes in magnetic recording. In MR microscopy, an MR head is raster scanned while in physical contact with a magnetic sample (e.g., hard disk media, tape, or fine magnetic particles). By plotting the MR resistance as a function of position, a very high resolution (on the order of. 1 x 1.0 mum) magnetic image of the sample is constructed. This case study focuses on characterizing the head sensitivity function (HSF), which depends on the physical dimensions and the magnetic properties of the MR head. These sensitivity functions are of great practical interest because they ultimately relate to the head's performance in a high-density data storage environment. Estimating the HSF requires a deconvolution that has features that prevent the problem from being straightforward: both the HSF and the source being scanned are unknown, and there is a substantial amount of correlated noise in the scanned image. We take a Bayesian approach to model and estimate the HSF, while accounting for noise and other nuisance effects such as thermal drift. Besides yielding a point estimate, which is a fairly difficult task here, this approach also quantifies uncertainty so we can assess whether certain features of the estimated head sensitivity function appear to be genuine.