A method for infrared image segmentation based on fuzzy kernel clustering using spatial relation was proposed.Firstly,the image was mapped into feature space.The fuzzy clustering was performed in this space.In order to suppress the noise,the part spatial information was introduced.And the classification reliability of pixel which was improved the degree of membership function,was used to analysis the classification rationality in the whole image.The classification reliability of pixel gray and the classification reliability of pixel distance were included in the the classification reliability of pixel.The experimental results show the infrared image can be segment well by fuzzy kernek clustering using the part spatial information and the whole spatial information.