Automatic, High-accuracy image registration in confocal microscopy

We proposed a High-accuracy image registration method of confocal microscopy for a large field of view and high resolution. The spatial information (edge information) and the entropy correlation coefficient have been both taken into account for higher accuracy of registration. The edge information is introduced to calculate the normalization correlation coefficient of the image. Then the normalization correlation coefficient and the entropy correlation coefficient of the original image have been used to improve the proposed similarity measures, the normalized mutual information with edge information (called EMI). Meanwhile, a parallel particle swarm optimization (pa-PSO) with the idea of conditional initialization and parallel cooperation is developed to speed up the convergence rate and further reduce the mismatch. Experiments verified that the registration accuracy can be up to 0.2 pixel and has better robustness to the noise.