Fast Estimation of Downsampling Factor for Biomedical Image Registration

Abstract

An approach to fasten medical image registration algorithms is suggested. It is based on the preliminary estimation the possible downsampling factor before the registration. The estimation algorithm uses fast bidirectional empirical mode decomposition. An analysis and approvement of the method is performed by multiscale ridge analysis using retinal image database DRIVE, astrocyte images and images from Computed Tomography Emphysema Database. Proposed registration acceleration algorithm was tested for rigid registration methods with HeLa cells video data set.