基于图割的DCT编码图像分割算法  

DCT Coded Image Segmentation Algorithm Based on Graph Cuts

作　　者：刘毅[1,2] 汪加才[1] 江效尧[1] 孙怀江[2] 夏德深[2] 

机构地区：[1]南京审计学院信息与科学学院,南京210029 [2]南京理工大学计算机科学与技术学院,南京210094

出　　处：《系统仿真学报》2012年第9期1859-1863,共5页Journal of System Simulation

基　　金：国家自然科学基金(60805003)

摘　　要：图割是图像分割中经典有效的算法,针对其计算量较大、实时性能不佳,面对广泛使用的DCT编码图像,提出一种改进的交互式快速分割算法。该算法利用DCT编码图像中的DC系数生成DC低频图像,有效降低了GMM参数学习的训练样本数,结合DC系数与AC系数生成纹理特征,通过颜色与纹理特征的有机结合、局部自适应正则化参数的选取,改善了对纹理图像和细长型边界的分割能力。实验结果表明,算法效率得到了提高,分割效果得到了一定的改善。Graph Cuts algorithm is a classical and effective method for image segmentation. But it is less efficient due to its high computational complexity. An improved interactive image segmentation algorithm was proposed to deal with the widespread use of DCT coded image. The DC coefficients of DCT coded image were used to generate DC low-frequency image for estimating the GMM parameters, which greatly reduced the scale of training samples. Then the DC and AC coefficients were used to generate texture features. With the combination of color, texture features and local adaptive regularization parameter selection the effects of texture and thin boundary image segmentation were improved. The experiments show that the efficiency of the algorithm and segmentation result has been improved.

关 键 词：图割 交互式图像分割 高斯混合模型 DCT 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]