In the process of designing and analyzing the yarn‐dyed fabric, the yarn color pattern has an important effect on the appearance of the fabric. An automatic color pattern recognition method for yarn‐dyed fabric is proposed in this study. The proposed method uses the fabric images obtained from a high‐resolution digital camera image acquisition system. The local statistical texture features are used for yarn texture segmentation. The yarn color classification problem is then formulated in a research framework of multiregion fuzzy segmentation, which can be added auxiliary variables and solved efficiently by the fast dual projection algorithm. The color values of the yarn crossing points are calculated by the yarn color classification results. The locations of the yarn crossing points are detected by a lightness gradient projection method. Different kinds of fabrics are tested in the experiments. Experiments on 14 actual fabrics show that the approach proposed in this study is effective for classifying yarn color and extracting the yarn color pattern in yarn‐dyed fabric.