In scalar volume visualization, features of interest are usually the boundary areas between homogeneous materials, and there have been many metrics to identify the boundary regions, such as the magnitude of the gradient and the scale values (LH values) of homogeneous materials. In this paper, we design a novel thickness-based transfer function to classify the boundary features according to their thickness, which is defined as the length of the integration path through the boundary in the derived gradient field. In combination with traditional scalar or LH transfer functions, the thickness-based transfer functions could further identify various features that could not be classified through these transfer functions solely. We performed experiments on various volume data sets to validate the effectiveness of our method, and the results showed that our proposed thickness-based transfer function could better improve classification of the boundary regions and had practical applications in medical and engineering fields.