Saadia Ayesha, Rashdi Adnan
Department of Computer Software Engineering, Military College of Signals, National University of Sciences and Technology (NUST), Islamabad, Pakistan.
Department of Electrical Engineering, Military College of Signals, National University of Sciences and Technology (NUST), Islamabad, Pakistan.
Comput Methods Programs Biomed. 2016 Dec;137:65-75. doi: 10.1016/j.cmpb.2016.09.006. Epub 2016 Sep 14.
Ultrasound is widely used for imaging due to its cost effectiveness and safety feature. However, ultrasound images are inherently corrupted with speckle noise which severely affects the quality of these images and create difficulty for physicians in diagnosis. To get maximum benefit from ultrasound imaging, image denoising is an essential requirement.
To perform image denoising, a two stage methodology using fuzzy weighted mean and fractional integration filter has been proposed in this research work. In stage-1, image pixels are processed by applying a 3 × 3 window around each pixel and fuzzy logic is used to assign weights to the pixels in each window, replacing central pixel of the window with weighted mean of all neighboring pixels present in the same window. Noise suppression is achieved by assigning weights to the pixels while preserving edges and other important features of an image. In stage-2, the resultant image is further improved by fractional order integration filter.
Effectiveness of the proposed methodology has been analyzed for standard test images artificially corrupted with speckle noise and real ultrasound B-mode images. Results of the proposed technique have been compared with different state-of-the-art techniques including Lsmv, Wiener, Geometric filter, Bilateral, Non-local means, Wavelet, Perona et al., Total variation (TV), Global Adaptive Fractional Integral Algorithm (GAFIA) and Improved Fractional Order Differential (IFD) model. Comparison has been done on quantitative and qualitative basis. For quantitative analysis different metrics like Peak Signal to Noise Ratio (PSNR), Speckle Suppression Index (SSI), Structural Similarity (SSIM), Edge Preservation Index (β) and Correlation Coefficient (ρ) have been used. Simulations have been done using Matlab. Simulation results of artificially corrupted standard test images and two real Echocardiographic images reveal that the proposed method outperforms existing image denoising techniques reported in the literature.
The proposed method for denoising of Echocardiographic images is effective in noise suppression/removal. It not only removes noise from an image but also preserves edges and other important structure.
超声因其成本效益和安全性而被广泛用于成像。然而，超声图像本身存在斑点噪声，这严重影响了这些图像的质量，并给医生的诊断带来困难。为了从超声成像中获得最大益处，图像去噪是一项基本要求。
为了进行图像去噪，本研究工作提出了一种使用模糊加权均值和分数积分滤波器的两阶段方法。在第一阶段，通过在每个像素周围应用一个3×3窗口来处理图像像素，并使用模糊逻辑为每个窗口中的像素分配权重，用同一窗口中所有相邻像素的加权均值替换窗口的中心像素。通过在保留图像边缘和其他重要特征的同时为像素分配权重来实现噪声抑制。在第二阶段，通过分数阶积分滤波器进一步改善所得图像。
针对人为添加斑点噪声的标准测试图像和真实超声B模式图像，分析了所提方法的有效性。将所提技术的结果与不同的先进技术进行了比较，包括Lsmv、维纳、几何滤波器、双边、非局部均值、小波、佩罗纳等人的方法、总变分（TV）、全局自适应分数积分算法（GAFIA）和改进分数阶微分（IFD）模型。从定量和定性两个方面进行了比较。对于定量分析，使用了不同的指标，如峰值信噪比（PSNR）、斑点抑制指数（SSI）、结构相似性（SSIM）、边缘保留指数（β）和相关系数（ρ）。使用Matlab进行了仿真。人为添加噪声的标准测试图像和两幅真实超声心动图图像的仿真结果表明，所提方法优于文献中报道的现有图像去噪技术。
所提的超声心动图图像去噪方法在噪声抑制/去除方面是有效的。它不仅能去除图像中的噪声，还能保留边缘和其他重要结构。
