This paper presents a new color-based approach for automated segmentation and classification of tumor tissues from microscopic images. The method comprises three stages: (1) color normalization to reduce the quality variation of tissue image within samples from individual subjects or from different subjects; (2) automatic sampling from tissue image to eliminate tedious and time-consuming steps; and (3) principal component analysis (PCA) to characterize color features in accordance with a standard set of training data. We evaluate the algorithm by comparing the performance of the proposed fully-automated method against semi-automated procedures. Experimental studies show consist agreement between the two methods. Thus, the proposed algorithm provides an effective tool for evaluating oral cancer images. It can also be applied to other microscopic images prepared with the same type of tissue staining.