结合颜色与纹理的JPEG图像快速分割  

Combination of Color and Texture for Fast JPEG Image Segmentation

作　　者：刘毅[1,2] 荆霞[2] 孙怀江[1] 夏德深[1] 

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

出　　处：《小型微型计算机系统》2013年第2期413-416,共4页Journal of Chinese Computer Systems

基　　金：中国数字化虚拟人切片图像分割研究(60805003)资助

摘　　要：图割方法是图像分割中的有效方法,为改善JPEG图像的分割效果,提高分割效率,提出将图割理论与JPEG图像的DCT系数相结合的方法.该方法以基于图割理论的GrabCut算法为基础,采用其交互简单的特点,为克服GrabCut算法基于全部像素迭代估计参数而导致算法效率低下,对高分辨率JPEG图像分割实时性差的不足,利用JPEG图像中的DC系数生成DC图像(低频图像)作为迭代估计GMM参数的样本点,缩小问题处理的规模,提高分割效率,提取水平、垂直、对角方向的AC系数作为纹理特征,颜色与纹理相结合,改善分割效果.实验结果表明,分割效果得到了改善,算法效率得到了显著提高.Graph cuts algorithm is an effective method for image segmentation. To improve the effect and efficiency of JPEG image segmentation, an approach of combining graph cut algorithm and DCT coefficients of JPEG image is proposed in this paper. On the basis of GrabCut algorithm, this approach makes use of its simple interactive mode and overcomes its shortcomings of inefficiency in estimating parameters based on the whole pixels, and poor real-time performance of segmentation for high-resolution JPEG image. The DC coefficients of JPEG image are used to generate DC image （ low-frequency image ） for estimating the GMM parameters effi- ciently and the scale of the problem is reduced. The horizontal, vertical and diagonal directions of AC coefficients are used as image texture features to combine with color features for improving the effect of segmentation. The experiments show that the proposed algo- rithm can get better segmentation results with a significant efficiency improvement.

关 键 词：图割 GRABCUT 高斯混合模型 JPEG DC系数 AC系数 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]