The desire to view smaller and smaller attributes within biological specimens means that confocal microscopes are often used at the limit of their resolution. For quantitative analysis of smaller sized attributes, and as a necessary preprocessing stage for automatic recognition and classification of objects it is essential that confocal images are restored. A fast new hybrid statistical restoration algorithm is presented which makes use of deterministic methodology to speed up optimisation of the posterior probability. Additionally, a prior probability model based on the Bayesian classifier is proposed. Restorations of real confocal image data using the above technique and prior are presented and discussed. Quantative analysis of the improvement gained through our hybrid approach is also presented.