Extensions of three single-channel segmentation algorithms to multi-channel operation are described. Algorithm performance is illustrated using multi-temporal ERS-1 images of an agricultural scene. It is shown that full multi-channel segmentation performed better than an approach based on segmenting channels separately and then recombining information. A multi-channel segmentor (RWSEG) based on segment growing constrained by edge detection was shown to perform better than a multichannel segmentor based on segment merging (MUM). However, in terms of preserving small scale detail neither preformed as well as a multi-channel filter (GAMANN) based on simulated annealing. It is illustrated how multi-channel segmentation may be used for identifying structural change and measuring radiometric change. With RWSEG, the choice of an RMS measure for combining information from different channels into a single parameter is somewhat arbitrary, its primary attraction being simplicity. More optimal methods probably exist and should be investigated.