The fracture surfaces have self-similar properties and are mostly expressed in the form of two-dimensional digital image. In this paper, a pixel-covering method is applied and an improvement is made to research the fractal characterization of the steel impact fracture surface. The results show that taking boundary to the feature parts of the gray image can increase the accuracy of the fractal dimension. The linearity of the fractal dimension curves of impact fractures is obvious, indicating the impact fractures have fractal properties and the pixel-covering method can describe it. Fractal dimension can be a parameter reflecting the roughness of the impact fracture surface. Rougher the surface is, the higher is the fractal dimension. There is a positive correlation between the fractal dimension of the impact fracture and the toughness of the material. It would be possible to establish a quantitative correlation between fractal dimension, surface roughness, impact toughness, and fracture mechanism, which presents a good potential to material and failure analysis of material. (C) 2012 Elsevier B.V. All rights reserved.