基于多值样条拟合曲线搜索的医学图像分割

针对肾组织切片图像的复杂特点,提出了一种基于拟合曲线搜索的肾小球分割方法,克服了边界增强方法中由于边界大量缺损而导致的肾小球区域提取不全的困难。首先在小波低频区域使用LOG滤波器获得含噪声较少的边界二值图像,在该二值图像中采用三次样条拟合方法构造闭合曲线,然后通过遗传算法在小波构造的低分辨率图像中搜索边界的最佳拟合曲线,在高分辨率图像中对该曲线进行精细化调整。实践证明了这一方法的正确性和有效性。

Medical image segmentation based on searching multi-valued spline fitting curve

A glomerulus segmentation method based on searching fitting curve is proposed by considering complex characteristics of the kidney-tissue image,overcoming edge enhancement method's shortcoming which resulted in extracted edge mass lack and thus glomerulus region can't be extracted completely.Firstly,a LOG filter is applied to the low-resolution image,which corresponds to low frequency sub band after wavelet transformation,so that rough edge information can be obtained.After labeling to remove the noises and thinning,a genetic algorithm is used to search close flitting curve by constructed a parametric equation based on cubic spline interpolation formula.Secondly,more detailed edge information can be acquired through applying LOG filter to the original resolution image.Then an elaborate adjustment is performed to resist the influence caused by other tissue noises. Practice has shown the correctness and effectiveness of this method.