In the literature on computer vision, very few contour detection algorithms are designed to deal with color images. In this paper, we present the multispectral contour detection algorithm (MSCDA) which is designed to process multispectral digital images as well as monochromatic ones. The MSCDA employs a bidimensional matrix of processing modules. The structure of a processing module is biologically plausible in that it consists of a bank of oriented filters. Each filter is a multispectral processing element (MSPE). A MSPE computes a contrast strength value locally from a receptive field characterized by specific orientation, shape, and size. The contrast strength value is a combination of an intensity contrast value with a chromatic contrast value, which are computed separately. Intensity contrast assesses the contrast due to local change in light energy, while chromatic contrast measures the contrast generated by local change in chromatic components. Even- and odd-symmetric MSPE pairs cooperate to extract a combined contrast strength value locally. Each processing module extracts one maximum combined contrast strength response from its bank of MSPEs. The maximum values of the combined contrast strength, provided by the grid of processing modules, form the contrast image. The contour candidate and the contour pixels can be extracted from the contrast image according to a strategy which is developed through simulations on 1-D and 2-D data sets. The MSCDA is compared with existing contour detection algorithms theoretically, and experimental results are shown. The MSCDA accounts for several psychophysical effects which are related to the mammalian visual system and may provide new insights into the understanding of the operational schemes employed by the visual cortex in combining energy, color and texture information for shape detection.