一种基于系数分类的CBCT图像去噪新方法

A New CBCT Denoising Method Based on Coefficient Classification

去噪是医学图像处理的一个重要研究课题。本文提出一种快速易实现的CBCT去噪新方法:利用二进小波变换将CBCT图像变换到小波域中,根据锥形影响域(COI)内的尺度间小波系数幅值和之间的比值,将小波系数分为两类,用基于方向窗的维纳滤波对不同类别的小波系数做不同的去噪处理,并提出一种更适合CBCT图像的噪声方差估计方法。对测试图像与临床CBCT图像的去噪结果表明,新方法所得去噪结果优于小波阈值去噪结果,能够在保留重要诊断细节的同时,有效地抑制CBCT图像中的噪声,为临床CBCT图像的实时去噪提供了一种新手段。

Denoising is an important issue for medical image processing. In this paper,a fast CBCT denoising method was proposed：CBCT images were transformed into wavelet domain with dyadic wavelet transform. According to the inter-scale relationship of wavelet coefficient magnitude sum in cone of influence （COI）,wavelet coefficients were classified into two categories,then different types of coefficient were denoised by different wiener filtering based on direction window at all levels,and a new noise variation estimating method more suitable for CBCT images was proposed. Experimental results of a test image and a clinical CBCT image show that this algorithm is superior to the conventional method for wavelet shrinkage denoising. This algorithm can suppress noise in CBCT images effectively and keep up the important structure details for diagnosis,thus providing a new approach for real-time denoising clinical CBCT images.