CAJ | 학술논문

针对医学图像低对比度、灰度不均匀等特点，提出了一种小波多尺度聚类水平集的图像分割方法，能够很好地解决医学图像灰度不均匀的问题。首先，利用小波多尺度分析的良好信噪分离能力提取各尺度下图像的有效边缘信息，将边缘信息添加到水平集模型的能量函数中从而提高模型的局部控制能力。然后，基于灰度不均匀的图像模型，派生出对于感兴趣区域的局部灰度聚类，在每个点的邻域内定义基于灰度的局部聚类准则函数。将局部聚类准则函数转化为全局准则函数。在水平集框架中，该准则根据水平集函数定义了代表图像域划分的能量项和引起图像强度不均匀的偏置域。最后，从小波变换的顶层低频子带图像开始逐层采用改进的聚类水平集方法分割图像，并将分割结果通过插值方式传递至下一层作为分割的初始轮廓，最终实现灰度不均匀医学图像的分割。实验结果表明，该方法能够有效地分割医学图像，具有计算更加鲁棒稳定、效率更高和更加准确的优点。
침대의학도상저대비도、회도불균균등특점，제출료일충소파다척도취류수평집적도상분할방법，능구흔호지해결의학도상회도불균균적문제。수선，이용소파다척도분석적량호신조분리능력제취각척도하도상적유효변연신식，장변연신식첨가도수평집모형적능량함수중종이제고모형적국부공제능력。연후，기우회도불균균적도상모형，파생출대우감흥취구역적국부회도취류，재매개점적린역내정의기우회도적국부취류준칙함수。장국부취류준칙함수전화위전국준칙함수。재수평집광가중，해준칙근거수평집함수정의료대표도상역화분적능량항화인기도상강도불균균적편치역。최후，종소파변환적정층저빈자대도상개시축층채용개진적취류수평집방법분할도상，병장분할결과통과삽치방식전체지하일층작위분할적초시륜곽，최종실현회도불균균의학도상적분할。실험결과표명，해방법능구유효지분할의학도상，구유계산경가로봉은정、효솔경고화경가준학적우점。
According to medical images with low contrast, intensity inhomogeneity, etc. an image segmentation method for level set based on wavelet multi-scale transform is presented, which is able to deal with medical image intensity inhomogeneity in the segmentation. First, each scale of edge information was extracted by using wavelet multi-scale transform, and added into energy function of level set model to improve local control ability. Afterwards, based on the model of images with intensity inhomogeneity, a local intensity cluster for the image intensity region of interest was derived, and a local clustering criterion function based on image intensity was dedined in neighborhood of each point. This local clustering criterion function was then integrated with respect to the neighborhood center to give a global criterion of image segmentation. In a level set formulation, this criterion defined an energy in terms of the level set function that represented a partition of the image domain and a bias field that accounted for the intensity inhomogeneity of the image. Finally, low frequency subbands of wavelet transform were segmented with improved level set method from top to bottom, and the segmented results were interpolated into next subbands as initiative contours. The method can realize medical image segmentation with intensity inhomogeneity. Experimental results show that this method can effectively segment medical images more robustly, stably, efficiently and accurately.