一种基于小波变换的低剂量 CT匹配插值方法

摘要

Interpolation of medical cross-sectional images is an important means for improving the quality of 3D reconstruction images . Low-dose computed tomography ( CT ) scanning results in increased X ray quantum noise and reduced entire resolution .In light of this problem, we present a matching interpolation method which is based on wavelet domain noise suppression .In the algorithm, after the tomographic images have been transformed to wavelet domain , for high-frequency sub-images the soft thresholding method is employed to suppress the noise first , and then the linear interpolation is made to obtain high-frequency sub-images with interpolation .For low-frequency sub-images the matching interpolation is used to obtain fairly accurate low-frequency sub-image of the interpolated image .Finally, the interpolated image is reconstructed by inverse wavelet transform based on the processed sub -images.Experimental results demonstrate that the proposed method can effectively suppress the noise and well keep the detailed information such as boundaries of the images .Comparative experiments of 3D reconstruction also show that by applying the proposed interpolation method , 3D reconstruction images have good results in reducing the noise and eliminating the terracing effect .%层间插值是提高三维重建图像质量的一个重要手段。针对低剂量CT扫描导致X线量子噪声增加和整体分辨率下降的问题，提出一种小波域抑制噪声的匹配插值算法。该算法将断层图像变换到小波域后，对高频子图先采用软阈值法进行噪声抑制再进行线性插值获取插值后的高频子图。对低频子图利用匹配插值获取较精确的插值图像的低频子图。最后，基于处理过的子图通过小波逆变换得到插值图像。实验结果表明，提议算法能有效抑制噪声和较好地保持图像边缘等细节信息。三维重建的比较实验也表明，利用提议方法后的三维重建图像在噪声减少及去除梯田效应方面有较好效果。