Different image-processing methods were used to count seed coat fragments (SCF) in cotton fiber webs. The results obtained showed that, similar to SCF detection on yarn, it is possible to use this technique to predict the "SCF potential" of lines during varietal improvement programs. The last method was tested on 440 images derived from carded cottons of various geographic origins. Counts obtained automatically by the analysis of images acquired by a camera were similar to SCF counts made visually by an "expert"; the relationship between the two counts being: square root of the automated count = 0.867 * square root of the visual count + 0.393 (r = 0.957).