Gradient-Domain Fusion for Color Correction in Large EM Image Stacks

Abstract

We propose a new gradient-domain technique for processing registered EM image stacks to remove inter-image discontinuities while preserving intra-image detail. To this end, we process the image stack by first performing anisotropic smoothing along the slice axis and then solving a Poisson equation within each slice to re-introduce the detail. The final image stack is continuous across the slice axis and maintains sharp details within each slice. Adapting existing out-of-core techniques for solving the linear system, we describe a parallel algorithm with time complexity that is linear in the size of the data and space complexity that is sub-linear, allowing us to process datasets as large as five teravoxels with a 600 MB memory footprint.