In order to accurately segment the coarse aggregate adhesion images in the CT X-ray asphalt mixture slice image during the classification of materials, an improved morphological multiscale algorithm with structural element radius of 1, 2, 3 and 4 was introduced to study the segmentation of the coarse aggregate adhesion images, respectively. By judging the number of the segmentation lines, the segmentation image with the maximum number of segmentation lines and minimal structural element was identified as the final segmentation. Then image segmentation was completed by overlapping the independent particle image and the segmented adhesion images. Study of effectiveness and accuracy were carried out to evaluate the improved algorithm. The test results showed that the improved morphological multiscale algorithm not only effectively separated the adhesion images of asphalt mixture X-ray CT slices, but also effectively reduced the over-segmentation and less-segmentation problems. The segmentation remains significantly higher in effectiveness and accuracy through this method which will reduce the difficulty of numerical modeling of the sample.