Defects connotative feature information in the X-ray image data of log with rotten knot was studied. Adopting box counting method detected log with rotten knot image edges effectively. Describe the irregular degree of image in quantity by using the value of fractal dimension to confirm the defects position. There were differences in fractal dimension between the normal regions and the edges. The values of normal regions in log image were about 2.0783, while that of the edges were between 1.4000 and 1.9000. The results show that fractal dimension was useful to numerically detect the defects of log with rotten knot image. This study provides a new method for edge detection of X-ray log image.