CAJ | 학술논문

经典的分水岭算法存在过分割问题,文中针对图像分割提出了一种基于预处理的改进算法.目的是为了抑制分割过程中的过分割现象,缓解过分割问题.首先对图像进行应用数学形态学去噪,滤波.再求取梯度图像,然后根据梯度图像局部极小值的综合信息自适应地提取内部标记,再进行距离分水岭变换提取外部标记.并以提取的标记为依据,对梯度图像进行修正,最后对修正后的梯度图像实施分水岭算法.实验结果表明:与传统分水岭算法相比,本算法能较好地缓解过分割问题.
경전적분수령산법존재과분할문제,문중침대도상분할제출료일충기우예처리적개진산법.목적시위료억제분할과정중적과분할현상,완해과분할문제.수선대도상진행응용수학형태학거조,려파.재구취제도도상,연후근거제도도상국부겁소치적종합신식자괄응지제취내부표기,재진행거리분수령변환제취외부표기.병이제취적표기위의거,대제도도상진행수정,최후대수정후적제도도상실시분수령산법.실험결과표명:여전통분수령산법상비,본산법능교호지완해과분할문제.
The over-segmentation is a problem of classical watershed algorithm,it presents an improved watershed algorithm based on pre-processing for image segmentation. The purpose is to suppress over-segmentation phenomena in the process of segmentation,alleviate the over-segmentation problem. Firstly,the morphological is used to denoise and filter. Then calculate the gradient image,subsequently, adaptively extract internal marker according to the integrated information of gradient image local minimum value,then carry on the dis-tance watershed transformation to extract external marker. Based on the marker,correct the gradient image,finally,the watershed algo-rithm is used to gradient image corrected. The result shows that compared with traditional watershed algorithm,this method can effectively remit the over-segmentation problem.