In order to better describe the complexity of the internal image quantitatively, the image complexity research method is proposed based on neural network, the method from three aspects to describe the complexity of the image, which is texture and edge information and significant area. The neural network technology is adopted to establish the image complexity and mathematical evaluation model between the various indexes. And the index weight values are obtained by the training for the neural network learning. The verification results show that the evaluation model is able to quantitatively describe the complexity of the internal image truly, and the experimental results obtained was consistent with human visual perception, and its feasibility in the algorithm model is verified.