提出了基于经验模态分解(简称EMD)方法复合材料X射线图像内部微小缺陷自动提取算法。该算法利用信息熵结合峰值信噪比，合理控制EMD分解层数，实现缺陷特征最优提取。通过试验结合图像分割，有效地分割出内部微小缺陷。相比于传统方法，该方法可检出由材料结构和图像噪声造成的微小缺陷以及所有方向和类型的缺陷，且可提取材料的其余特征值。

A method based on EMD(Empirical Modal Decomposition) for X-ray detection and extraction for tiny flaws in composites was presented. This method took advantage of information entropy and PSNR (peak signal to noise ratio) to control EMD decomposition order and realize the optimum extraction of defect characteristics. By using experiment and image segmentation, it could segment tiny flaws effectively. Comparison with traditional methods, it could extract tiny flaws from material structure and image noise, detect all types of tiny flaws in any direction and extract other material characteristic which lack of commonality and had specific applications.