CT图像局域阴影特征小波阈值提取方法

Wavelet Threshold Extraction Method of Local Shadow Feature of CT Image

利用当前方法对CT图像局域阴影特征提取时未对CT图像进行预处理,存在峰值信噪比低、ROC曲线低和特征分离度低的问题。提出CT图像局域阴影特征小波阈值提取方法,该方法确定小波阈值后对CT图像进行去噪,同时利用线性、分段线性以及非线性三种灰度变换方式以此增强CT图像分辨率,以便更加清晰地分辨CT图像局域阴影,并利用二维Daubechies小波分解CT图像局域阴影,将其分成若干个小图像后利用概率密度分布求解特征值,最终结合所有图像的特征值实现CT图像局域阴影特征提取。实验结果表明,所提方法的峰值信噪比高、ROC曲线高和特征分离度高。

The current method does not preprocess the CT image when extracting the local shadow feature of CT image, which has the problems of low peak signal-to-noise ratio, low ROC curve and low feature separation. A wavelet threshold extraction method of local shadow feature of CT image is proposed. According to the method, the CT image was de-noising after the wavelet threshold was determined. In order to distinguish the local shadows of CT images more clearly, linear, piecewise linear and nonlinear were used to enhance the resolution of CT images. The two-dimensional Daubechies wavelet was adopted to decompose the local shadow of CT image, and then it was divided into several small images, and the probability density distribution was applied to solve the eigenvalue. Finally, according to the feature values of all images, the feature extraction of local shadow in CT images was realized. The experimental results show that the method has high peak signal-to-noise ratio, ROC curve and feature separation.