An automatic detection system for gynecologic cervical cancer cells from Papanicolaou-stained smears is proposed. The system is based on multispectral scanning, with smear images provided as three monochromatic images by illumination with light of 610-, 535- and 470-nm wavelengths. Scene segmentation in the system is performed by classifying all two-dimensional pixels defined at all image points in the red and green images into one of three classes (background, cytoplasm and nucleus) by means of a maximum-likelihood classifier. The statistical parameters for the classifier are determined under certain assumptions from a two-dimensional histogram, which indicates the bivariate distribution of the optical densities in the red and green images. The nuclear characteristic parameters extracted from individual nuclei by the system are (1) nuclear area, (2) nuclear diameter, (3) nuclear area coefficient and (4) nuclear density vector, which consists of the average optical densities of the red, green and blue images of the nuclear subregion. The system was tested on 34 scenes containing cellular clumps of different cell types. The results of the segmentation of images and the performance of a primitive logic for the detection of malignant nuclei proved the usefulness of the system and the advantages of using multispectral images in the segmentation and detection procedures. The system can detect not only a free-lying cancer cell but also a malignant cell in a clump of cells.