<p>In the process of pencil drawing simulations using Line Integral Convolution (LIC), noise plays a crucial role. This study proposed a novel training-free hybrid noise model by reversing the monotonicity between the variance and the grayscale values of the input image in the existing hybrid noise model. It can address the issue that the existing hybrid noise model cannot generate clear and visible strokes in the bright region. The grayscale values of the input image are allowed to influence the grayscale values of the noise model in the proposed improved hybrid noise model, which can ultimately affect the grayscale values of the generated pencil drawing. Therefore, the problem of poor simulation of the hatching graduation in stroke intensity and stroke density is solved. Experiments have demonstrated that the proposed improved hybrid noise model can simulate clear and visible strokes and performs better in simulating the hatching graduation in stroke intensity and stroke density.</p>