Thyroid diseases are prevalent among endocrine diseases. Observation and examination of histological tissue images can help in understanding the cause and pathogenesis of the tumor. The aim of this study was to quantify the histological image features of microscopic thyroid images in order to classify varying tissue types. Five typical histological thyroid tissues were characterized using seven image features including hue, brightness, standard deviation of brightness, entropy, energy, regularity, and fractal analysis. Statistical stepwise selection and multiple discriminant analysis were then used to classify the features. The results show all of the features are significant and our algorithm has the capability of differentiating histological tissue types. The algorithm is applied utilizing quad-tree based region splitting methods to segment the tissue regions from the heterogeneous microscopic image. The preliminary results show the system has good performance for tissue segmentation.