MR脑图分割算法研究

A Segmentation Algorithm Based on Labeling Watershed and Wavelet

在图像处理精确分割有用图像的研究中,目前MR脑图分割中边缘定位不准和分割精度不高。为了解决上述问题,提出将标记分水岭算法和小波变换相结合的MR脑图分割算法。方法首先利用小波变换的多尺度思想对图像进行多分辨率分解和去噪,然后在低分辨率图上应用标记分水岭算法,并对标记好的低分辨率图像进行中值滤波,最后利用小波逆变换将低分辨率图像反建为高分辨率图像。对方法和现有的方法进行仿真,结果表明可以准确地定位图像边缘,有效地提高了分割精度,效果明显优于传统的分割算法。

With regard to the low accuracy in edge location and segmentation of current MRI segmentation algorithm,an algorithm of MR brain segmentation combining labeling watershed and wavelet transformation was presented.Based on the theory of multi-scale,this method firstly deconstructed and denoised the image in multi resolution,and then segmented the image using labeling watershed in low resolution,and then median-filtered the image,at last,constructed the high-resolution image from low-resolution image.By comparing this method with current ones through simulation,it is proved that this method can precisely locate the image edge in high accuracy,which performs much better than current algorithms.