探讨了基于分数阶微分增强和独立分量分析在红外图像边缘提取上的应用。首先,从数字图像分数阶微掩模及其运算规则出发对红外图像进行增强。然后,基于信息最大化算法对自然图像进行迭代训练,得到ICA所需的基函数。最后,运用独立分量算法对增强后的红外图像进行边缘提取。实验结果表明,分数阶微分在增强红外图像灰度变化不大的平滑区域中的边缘特征效果明显,而ICA算法做边缘提取即使在有噪声的情况下也能较好地提取红外图像的边缘特征。
The application of a new method of infrared image edge extraction based on fractional derivative approach and ICA is discussed in this paper. First,we can get the image enhanced by fractional derivative approach. Then, it can get the basis functions of natural images by using the information maximization algorithm of ICA. The last, we use ICA to extract the edge of infrared image. The experimental result shows that the fractional derivative approach can enhance the smooth area of image and the ICA has the great effect on extracting the edge of infrared image even if there is any noise exist.