This paper introduces a method for automatic analysis of wood defects in CT images through artificial neural networks （ANN）. Experiments on European ash （Fraxinus excelsior） showed that the ANN of nodes in different hidden layers could accurately distinguish wood bark, knot, decay and clear logs, but it did not work well for fine splits. The application of the method is helpful to realize quick and automatic recognition of wood defects, and to draw an optimal sawing scheme.