基于小波变换与Snake模型的肺野图像分割方法

A LUNG FIELD IMAGE SEGMENTATION METHOD BASED ON WAVELET ANALYSIS AND SNAKE MODEL

针对X射线DR尘肺胸片受到尘肺病变多样性影响导致基于区域和边缘分割的方法无法获取肺野精确分割的问题,提出基于小波变换与Snake模型算法相结合的肺野分割方法。通过对X射线DR尘肺胸片特征进行分析,首先在空间域对目标图像的边缘进行线性变换增强,然后对图像进行基于小波的多尺度分解,并利用Canny算法对不同尺度的低频图像边缘进行提取和融合,实现自动获取肺野图像的初始轮廓。最后用GVF Snake模型算法对该初始轮廓进行修正优化,从而实现对肺野区域的精确分割。实验结果证明,所提方法具有较高的识别率和稳定的识别效果。

X-ray chest radiographs of DR pneumoconiosis is affected by the diversity of pneumoconiosis disease, which lead to the region-based and edge-based image segmentation algorithms not being able to make accurate segmentation on lung field region. In light of this problem, we propose a lung field segmentation algorithm which is based on the combination of wavelet transform and Snake model algorithm. By analysing the features of X-ray chest radiographs of DR pneumoeoniosis, the algorithm first enhances the edge of the target image through linear transform in spatial domain, then it decomposes the image in multi-scale with wavelet method. After that, the Canny algorithm is employed to extract and integrate the edges of the low frequency images with different scales to implement the automated acquisition of initial profiles of lung filed image. Finally, the initial profiles are further modified and optimised with gradient vectorflow filed （GVF） Snake algorithm so as to realise accurate segmentation on lung field region. Experimental results prove that the proposed method has a higher recognition rate and stable recognition effect.