This article presents a multi-spectral edge detection method that combines the best approaches from several sources. It uses all available multi-spectral information, by adding the magnitudes and directions of edges derived from edge detection in single bands, instead of just taking the maximum, which is commonly done. The addition is weighted on the edge direction, to remove noise and enhance the "major direction". The resulting image is then thinned using magnitude and direction information, together with hysteresis thresholding, so that weak edges can be preserved without adding too much noise. Small gaps in the edges are filled in. All edges between areas with different spectral characteristics are detected. The same technique could also be used for edge detection in multi-temporal images, or even multi-temporal/ multi-spectral images. The edge image can then be used for further analysis, such as field segmentation and subsequent classification.