The work aims to effectively remove the noise in the visual inspection system of tablet packaging, to enhance the image clarity and ensure that the later image segmentation and edge processing is carried out smoothly. For a large number of uncertain noises in the visual inspection images of tablets, an image filtering algorithm of the adaptive fuzzy neural network was proposed. A robust membership function was introduced in the fuzzy neural network structure, and the optimization training on the parameters of fuzzy neural network was carried out in the gradient descent method. The images with noise were filtered by the optimized network structure. The simulation results showed that the proposed algorithm could eliminate the noise of the tablet whiling preserving complete image edge and important details. The proposed filtering algorithm can effectively improve the tablet image clarity, and is of great significance for the later tablet image segmentation and edge processing.