摘要
基于图割理论的GrabCut算法具有全局最优性和结合多种知识的统一性,但其基于全部像素点的参数估计以及为达到一定分割精度采取的迭代求解模式,使算法效率大大降低。以GrabCut算法为基础,通过小波变换将图像分解,用分解后低频图像的像素点作为GMM参数迭代估计的样本点,减小了问题规模。实验结果表明,算法的效率得到较大提高。
GrabCut algorithm based on graph cuts has the global optimality and the unity of combining multiple knowledge. However, such algorithm is less efficient because it uses the whole pixels to initialize the GMM parameters and uses iterative algorithm to obtain exactitude.On the basis of GrabCut algorithm,this paper processes the image using wavelet transform,and then estimates the GMM parameters with low-frequency image＇s pixels,sharply decreases the problem scale.The experiments show that this method significantly improves the algorithm＇s efficiency.
出处
《计算机工程与应用》
CSCD
北大核心
2010年第33期215-217,共3页
Computer Engineering and Applications
基金
陕西省自然科学基金No.2005A12
陕西师范大学研究生培养创新基金(No.2008CXS025)~~
关键词
小波变换
图像分割
高斯混合模型
目标提取
wavelet transform
graph cuts
Gaussian Mixture Model（GMM）
object extraction
