A new image segmentation method based on discrete Tchebichef moments and quantum neural networks is presented. The Tchebichef moments in certain local window of each pixel in the image are computed and input to quantum neural network . Quantum neural networks , which use multilevel transfer function, have the inherent fuzzy characteristics. The point accommodates to the connatural uncertainty of fractional image data in image segmentation procession. Experiments confirm that the performance of our proposed method is more accurate and has less iterative times in comparison with the traditional segmentation methods based on Legendre moments and BP neutral networks.