基于超像素特征表示的图像前景背景分割算法

西安科技大学 计算机科学与技术学院,陕西 西安 710054

图像分割算法; 超像素提取; 线性分类器; 特征表示

Foreground and background segmentation based on superpiexel-level feature representation
XUE Ping

(College of Computer Science and Engineering,Xi'an University of Science and Technology,Xi'an 710054,China)

image segmentation; superpixel extraction; linear classifier; feature representation

DOI: 10.13800/j.cnki.xakjdxxb.2017.0520

备注

图像前景背景分割是图像处理中的关键技术,文中提出了基于超像素分类的二值分割算法。对于输入图像,首先采用超像素分割算法,将图像分割成多个保留边缘的封闭区域,即超像素; 对每一块超像素,考虑颜色和纹理,构造一种对光照和颜色较为鲁棒的特征,来消除同种物体在光照和颜色差异下的影响; 用所得特征训练分类器,判断每块超像素属于前景或背景; 最后将超像素分类结果作为初值用图分割的方法进行修正,得到最终的二值分割结果。实验结果显示算法能较好的完成前景背景分割的任务。此外,本算法易于和现有的分类算法相结合,具有较强的可移植性。

The foreground and background segmentation is an important technique in image processing.In this paper,a binary segmentation method is proposed based on the classification of superpixel.The input image is firstly divided into several superpixel to protect the edge of objects.For each superpixel,the color and texture are considered to extract the feature with robust for illumination and color,which can eliminate the influence of light and color.The feature vectors are further used to train a classification to classify the superpixel into foreground or background.Finally,the graph cut method is used to modify the class label of each pixel with the initialization of superpixel.The experiment result shows that the method can successfully extract the objects from the background.Moreover,this method is easy to be implemented since it can be combined with the classification technique directly.