摘要
针对传统Grabcut分割方法难以很好处理图像阴影部分及分割时间过长的问题,提出结合K-means聚类算法与Grabcut函数的方法改善以上问题。该方法通过直方图均衡化实现图像增强,然后利用K-means算法以二分类的形式对像素进行聚类,利用形态学处理填充孔洞,自动获取目标物体轮廓后,结合Grabcut算法实现图像二次分割,最后利用颜色判定改善图像阴影部分。实验结果表明,该方法可实现图像自动分割、节省大量时间,而且对阴影部分图像处理有明显改善,具有较高的准确性和高效性。
The traditional Grabcut segmentation method is unable to deal with some shadows and the segmentation time is too long. Aim-ing at these problems,the paper puts forward the combination of K means clustering algorithm and Grabcut function. The method realiz-es image enhancement by histogram equalization,then it uses the K-means algorithm to cluster the pixels in the form of two classifica-tions,fills the holes with morphological processing,automatically obtains the contour of the target object,and combines Grabcut algo-rithm to realize image quadratic segmentation. Finally,color determination is used to improve the shadow portion of the image. Experi-mental results show that this method can automatically segment images and save a lot of time. It also improves the shadow images signif-icantly and has high accuracy and efficiency.
作者
王告
俞申亮
巨志勇
马素萍
WANG Gao;YU Shen-liang;JU Zhi-yong;MA Su-ping(College of Photo-electricity and Computer Engineering,Shanghai University of Technology;Shanghai Harbor e-Logistics Software Co.,Ltd.,Shanghai 200093,China)
出处
《软件导刊》
2019年第6期171-175,2,共5页
Software Guide
