Electrophoresis is a process for identifying macromolecules such as proteins or nucleic acids in a substance by 2D spatial separation. The process results in a 2D image with dark spots over a light background indicating the presence of certain macromolecules. Due to the presence of noise, several samples of the same substance are usually obtained. A common task is to compare these samples to determine which spots appear in a the majority. Such analysis requires registration and segmentation. In this presentation, an automatic segmentation scheme for electrophoresis images is presented. Multiple samples are mutually registered and a single segmentation is obtained in which each spot is assigned a unique label. Markov random field methods are used, both for the registration and the segmentation. The registration algorithm allows for spots to be present in some images and not in others. The segmentation algorithm can correctly label spots that partially overlap.