基于频域快速解卷的血管提取算法

Vessel Extraction Algorithm Based on Fast Frequency-domain Deconvolution

为了方便识别新生血管,从而有效地对疾病进行诊断,所以需要将血管图像从背景图像中分割出来。文中提出一种在提高血管分辨率的基础上进行血管分割的算法。首先通过特定公式在梯度图像上对血管图像进行纹理增强,这是为了在之后的图像分割中,能识别细微的血管,使其不被忽略掉;然后对模糊图像进行频域上的快速解卷去模糊,消除成像仪器在成像过程中对图像清晰度的影响;由于图像在传输过程中会产生噪声,因此为了去除噪声对血管分割的影响,接着对血管图像进行了中值滤波操作;最后使用最大类间方差法来进行血管的分割操作,因为最大类间方差法可以有效地将图像前景和背景分割开。通过实验对比,直观上证明了该算法在血管分割中的有效性。

In order to identify new blood vessels, thus effectively diagnosing disease, it is necessary to segment the vascular image from the background image. A segmentation algorithm based on improved resolution of the vessel is proposed. First the texture is enhanced for blood vessel image through specific formula in the gradient image,in order to identify small blood vessels and not ignore it in the image segmentation later. Secondly,the image deconvolution in frequency domain is made to obtain a sharp image, eliminating the influence on image clarity of imaging instrument in the imaging process. Then so as to remove the influence of noise on the vessel segmentation, a filtering operation is carried on vascular image. Finally ,the maximum class variance method is used to segment vessels,because it can effectively distinguish the image foreground and background. The effectiveness of the algorithm in vessel segmentation is verified intuitively by experimental comparisons.