摘要
传统的Livewire算法仅通过目标和背景的亮度差异获得目标边缘,以人机交互方式完成目标分割。为提高目标分割的精度和效率,进一步引入了目标和背景之间的色彩差异、纹理差异等信息。新的Livewire算法计算像素的亮度、色彩、纹理的统计直方图梯度,用logistic回归合成图像的边缘特征,再应用Dijkstra最短路径算法来完成目标分割。分割测试结果表明:在多数图像分割中,多特征的Livewire算法在分割精度和交互效率上要优于仅使用亮度特征的传统方法。
Traditional Livewire image segmentation algorithms get object edges based on brightness differences of object and background, and achieve object segmentation by human computer interaction. In order to improve the accuracy and efficiency of segmentation, a new livewire algorithm is proposed in which the information such as color difference and texture difference between the object and the background are applied. The new algorithm computes the histogram gradients of brightness,color,and texture of pixels,which synthesizes the image edges by logistic regression and segments the objects by Dijkstra＇s shortest path algorithm. The segmentation experiments shows that the new livewire algorithm with multiple features has better accuracy and efficiency than the traditional ones which only use brightness in many segmentations.
出处
《广西大学学报（自然科学版）》
CAS
北大核心
2017年第2期742-747,共6页
Journal of Guangxi University（Natural Science Edition）
基金
国家自然科学基金资助项目(61572285,61272237)
湖北省水电工程智能视觉监测重点实验室2016年开放基金资助项目(2016KLA05)
关键词
图像处理
图像分割
边缘检测
image processing
image segmentation
edge detection
