Since rock fractures have a strong impact on radionuclide transport and retention, fracture edge detection is very important for rock engineering. A number of rock specimens were sampled, and then the rock images were acquired for detecting the fracture edges. Edge detection in color images could be seen as detecting changes in a vector field. The quaternion convolution at different scales is applied. In addition, to suppress image noise, the dot product was computed at adjacent scales. At the same time, the gray-level difference was applied to obtain the monochromatic edges. To merge the merits of the quaternion convolution and the gray-level difference, the two results are combined. In this way, the thinned edges can be obtained by using modulus maximum suppression. Experimental results show that the proposed algorithm is efficient and robust for rock fracture edge detection.