Eddy current measurements of cracks in titanium can be corrupted by the presence of spatially correlated grain noise generated by the material. This grain noise complicates both the detection and characterization of cracks in titanium. A promising solution is to model the grain noise as a Gaussian process; this model was successfully applied in the characterization problem (see [1]). Here, a grid search algorithm which makes use of the noise model is developed to detect cracks obscured by grain noise. As in [1], the inverse problem is solved in the Bayesian framework. The algorithm is applied successfully to both simulated and experimental data for cracks in specimens of Ti-6Al-4V.