The resolution of contour feature of images directly relies on the optimization and determination of the threshold value in image binaryzation process. However, the binaryzation threshold value of an image has turned out to be rather flexible as cross-section of Chinese herbal pieces varies dramatically in terms of complexity, evenness and shape. It is meaningful to develop a method to definitely choose a proper threshold value in order to portray the contour feature of cross-section texture images of Chinese herbal pieces clearly after the image binaryzation process when it comes to image pattern recognition. A method based on image complexity research was proposed to determine a proper value of binaryzation. In this method, the binaryzation threshold value varies with the variation of image complexity, while the image complexity was determined according to differences between one pixel and its neighborhoods of an image. The result showed that binaryzation images whose threshold value derived from the method proposed in this study were well able to depict the features of cross-section texture of Chinese herbal pieces.