It has been difficult to obtain better than moderate resolution in analysis of electron microscopic images of small, 2D crystals with variable lattice parameters, e.g., crystals of the channel VDAC generated by phospholipase treatment of outer mitochondrial membranes. We demonstrate that applying single-particle analysis methods to correlation-averaged images can lead to significant improvements in the attainable resolution. Application of a soft-edged fitted mask passing only the central unit cell, and excluding the positionally variable adjacent unit cells, allows improved alignment and more sensitive multivariate statistical analysis, needed to guide intelligent merging of data from different crystals.