In this paper，an adaptive bilateral filter is proposed to remove noises in 3D mesh models． Firstly，the feature detection of 3D mesh models is performed on the basis of the volume integral invariant，and the local feature strength is estimated． Then，the proposed filter adaptively adjusts the denoising parameters in the bilateral filter according to the local feature strength． Through the adaptive optimization of parameters，the proposed filter provides targeted denoising strategies for the regions with different feature strength，thus further improving the denoising performance． Experimental results reveal that，in comparison with common bilateral filters，the proposed adaptive bilateral filter achieves a superior performance in preserving the sharp features of 3D mesh models during the denoising，and the objective distance measure between the denoised mesh model and the original model is averagely reduced by 0. 0332．