基于 MAP 的超声图像分解去噪算法研究

摘要

超声图像中的斑点噪声，降低图像分辨率和对比度，不利于后续图像处理。本文基于最大后验概率（Maximum A Posteriori ，MAP）推导出一种新的超声图像分解算法，将原始超声图像分解为无散斑真实图像和散斑图像。使用六组不同的参数值，对 Field II 仿真的超声图像进行分解试验，得出算法中比例参数对分解结果的影响规律。用该方法分解三幅人体超声图像，得到的真实图像平滑性好，且能较好的保留细节和边缘。本文提出的分解算法可用于超声图像的去噪，且分解得到的真实图像和散斑图像可用于特征提取、图像分割和图像分类等。%The speckle noise in ultrasound images reduces image resolution and contrast ,and brings difficulty to subsequent image processing .Based on MAP (Maximum A Posteriori) ,the paper proposes an ultrasound image decomposition method ,to de-compose an observed image into a speckle-free true image and a speckle image .We use Field II to simulate an ultrasound image , and decompose it with the proposed method .By six tests with different parameters ,we study how the weighted parameters influence the decomposition results .Then we use this method to decompose three real ultrasound images .The true image we get by decompo-sition can be seen as the denoised image ,because it has good homogeneity ,and preserves details and edges well .Both the true image and the speckle can be used in feature extraction ,image segmentation ,classification and so on .