The optical characteristics and differences of rock components in the thin section were analyzed on the basis of images at sequential extinction angles under the cross-polarized microscope. A new methods for grain segmentation and pore identification were proposed. Taking full advantage of the modes of occurrence of grains and the complexity of crystal optical axis, the spatial arrangement of grains and the diversity of contact relationships, the authors carried out the correlation analysis of pixel scales on the extinction angle of thin sections. The grain segmentation and the pores identification were brought about by sensitive parameters, such as the mean and the standard deviation of the correlation coefficient. The results show that the gray level and RGB values of the pixel inside the grains are more correlated in the angular domain. At the edge of a grain and distribution area of fillings in pores, the correlation coefficient was lower, and the variance of correlation coefficient was obviously higher than that in the interior of grains. This method demonstrates the significance of information integrity in the angle domain from the extinction characteristics. The edge extracted by this method is clear, the skeleton of the pore structure can be characterized, and the accuracy of grain segmentation is better than that of Sobel and Canny.