光学相干断层扫描(OCT)作为眼科疾病的重要诊断手段之一, 具有巨大的发展潜力。OCT图像中含有大量的散斑噪声, 会影响图像病理信息的检测和提取。提出一种去除OCT图像散斑噪声的方法, 依据散斑噪声的乘性特性建立了乘性噪声模型。针对双域滤波中产生的伪吉布斯现象, 引入各向异性扩散获得引导图像, 代替了算法中的自身引导, 减少了迭代次数。结果显示, 所提方法提高了算法效率与图像质量, 算法运行时间大幅度下降, 图像峰值信噪比(PSNR)提高到36dB, 结构相似性(SSIM)和边缘保持系数(EPI)分别达到0.8和0.7左右, 有效抑制了OCT图像中的散斑噪声。

Optical Coherence Tomography (OCT) has great potential as one of the important diagnostic tools for ophthalmic diseases. The noise in the OCT image is mainly speckle noise, which will affect the detection and extraction of image pathological information. A method for removing speckle noise in OCT images is proposed. The multiplicative noise model is established for the multiplicative characteristics of speckle noise. For the Pseudo-Gibbs phenomenon caused by dual domain filtering algorithm in wavelet shrinkage in the frequency domain, an anisotropic diffusion algorithm is introduced in the preprocessing stage to obtain a guided image, which replaces the self-guided in the original algorithm and reduces the number of algorithm iterations. The results show that the proposed method improves the efficiency of the algorithm and the image quality. The running time of the algorithm is greatly reduced. The peak Signal-to-Noise Ratio (PSNR) of the image is 36dB, and the Structural Similarity Index (SSIM) and Edge Preservation Index (EPI) are 0.8 and 0.7 respectively. The result proves that the improved dual domain filtering algorithm can effectively suppress speckle noise in OCT images.