For effective detection of foreign fibers which have almost the same shape and color as the cotton fibers, an infrared spectral imaging method based on the absorption characteristic discrimination between cotton fibers and foreign fibers was proposed. According to the method, the relation between the absorption characteristic discrimination and the wavelength in the region of near infrared was analyzed, and the optimal band for discriminating several types of foreign fibers from cotton fibers was determined. The absorption characteristic of foreign fibers was transformed into the image feature in the infrared spectral imaging system. Furthermore, the foreign fibers were extracted form the background of lint by using adaptive image enhancement and binary image processing. The experimental result indicates that the image features of foreign fibers are obvious. And it provides an effective method to detect foreign fibers in lint.